Wireless communication device for correcting non-linearity of transmitter and method of operating same

Abstract

A wireless communication device for correcting nonlinearity of a transmitter and a method of operating the same, the method of operating the wireless communication device comprising: generating an IQ compensation value by performing IQ mismatch compensation on a first input signal using a linear calibration model, the linear calibration model being based on a neural network; generating a predistortion value by performing predistortion on the first input signal using a linear calibration model; generating a first calibration signal based on the IQ compensation value and the pre-distortion value; generating a calibrated first output signal by amplifying the first calibration signal based on the amplification coefficient; and training a linear calibration model based on the first input signal and the calibrated first output signal.

Description

[0001] This application is based on and claims the benefit of priority from Korean Patent Application No. 10-2023-0134362 filed on October 10, 2023 in the Korean Intellectual Property Office, the disclosure of which is incorporated herein in its entirety by reference. Technical Field

[0002] The inventive concept relates to a wireless communication device, and more particularly, to a wireless communication device for correcting nonlinearity of a transmitter and an operating method thereof. Background Art

[0003] Recently, with the rapid development of wired communication and wireless communication technologies and smart device-related technologies, the need to correct nonlinearity of a transmitter of a wireless communication device in a wireless communication system has increased.

[0004] Typically, a transmitter of a wireless communication device may include a power amplifier configured to generate a transmission signal by amplifying the strength of an input signal. Since the power amplifier has nonlinearity, the nonlinearity of the transmission signal relative to the input signal of a specific level or higher increases. In order to correct the nonlinearity of the power amplifier, methods such as crest factor reduction (CFR), predistortion, and IQ mismatch compensation may be used. However, for example, since a heuristic method is used to select appropriate coefficients to be used for predistortion, it may be difficult to select appropriate coefficients, resulting in performance degradation. Therefore, a method for effectively correcting the nonlinearity of a transmitter of a wireless communication device is needed. Summary of the invention

[0005] The inventive concept provides a wireless communication device having nonlinearity-corrected transmission performance by correcting the nonlinearity of a transmitter of the wireless communication device based on a neural network and an operation method thereof. According to an embodiment, a method of effectively correcting the nonlinearity of a transmitter of a wireless communication device is provided.

[0006] The technical challenges solved by the inventive concept are not limited to the above-mentioned technical challenges, and a person of ordinary skill in the art may clearly understand other unmentioned technical challenges from the following description.

[0007] According to one aspect of the inventive concept, a wireless communication device is provided, including: a first processing circuit, configured to generate a calibration signal by performing linear calibration on a first input signal using a linear calibration model, the linear calibration model being based on a first neural network; and a power amplifier, configured to generate a calibrated output signal by amplifying the calibration signal based on an amplification factor, the linear calibration model including a first output layer and a second processing circuit, the first output layer being configured to generate a first IQ compensation value and a second IQ compensation value by performing IQ mismatch compensation on the first input signal, the first IQ compensation value corresponding to a real part of the calibrated output signal, and the second IQ compensation value corresponding to an imaginary part of the calibrated output signal, the second processing circuit being configured to generate a first pre-distortion value and a second pre-distortion value by performing pre-distortion on the first input signal, the first pre-distortion value corresponding to the real part of the calibrated output signal, and the second pre-distortion value corresponding to the imaginary part of the calibrated output signal, and the first processing circuit being configured to generate the calibration signal based on the first IQ compensation value, the second IQ compensation value, the first pre-distortion value, and the second pre-distortion value.

[0008] According to another aspect of the inventive concept, there is provided an operating method of a wireless communication device, the method comprising: generating an IQ compensation value by performing IQ mismatch compensation on a first input signal by using a linear calibration model, the linear calibration model being based on a neural network; generating a predistortion value by performing predistortion on the first input signal by using the linear calibration model; generating a first calibration signal based on the IQ compensation value and the predistortion value; generating a calibrated first output signal by amplifying the first calibration signal based on an amplification factor; and training the linear calibration model based on the first input signal and the calibrated first output signal.

[0009] According to another aspect of the inventive concept, there is provided an operating method of a wireless communication device, the method comprising: generating an IQ compensation value by performing IQ mismatch compensation on a first input signal using a linear calibration model, the linear calibration model being based on a first neural network; generating a predistortion value by performing predistortion on the first input signal using the linear calibration model; generating a first calibration signal based on the IQ compensation value and the predistortion value; generating a calibrated first output signal by amplifying the first calibration signal based on an amplification factor; and training a power amplifier estimation model based on the first input signal and the first output signal to obtain a trained power amplifier estimation model, the power amplifier estimation model being based on a second neural network, and the first output signal being obtained by amplifying the first input signal based on the amplification factor. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Embodiments of the inventive concept will be more clearly understood from the following detailed description in conjunction with the accompanying drawings.

[0011] Figure 1is a block diagram showing a wireless communication system according to an embodiment.

[0012] Figure 2 is a block diagram showing a wireless communication device according to an embodiment.

[0013] Figure 3 is a block diagram illustrating a neural network applied to a linear calibration model according to an embodiment.

[0014] Figure 4 is a block diagram showing a wireless communication device according to an embodiment.

[0015] Figure 5 is a block diagram showing a wireless communication device according to an embodiment.

[0016] Figure 6 is a block diagram illustrating a linear calibration model according to an embodiment.

[0017] Figure 7 is a graph comparing an adjacent channel leakage ratio (ACLR) of an output signal generated by a wireless communication device according to the embodiment with an ACLR of an output signal according to a comparative example.

[0018] Figure 8 is a flowchart for describing an operation method of a wireless communication device according to an embodiment.

[0019] Fig. 9 is a flowchart for describing an operation method of a wireless communication device according to an embodiment.

[0020] Fig.10 is a block diagram showing a wireless communication device according to an embodiment. DETAILED DESCRIPTION

[0021] Figure 1 is a block diagram showing a wireless communication system WCS according to an embodiment.

[0022] Hereinafter, although an embodiment is described with reference to a wireless communication system WCS based on a New Radio (NR) network, the technical concept of the inventive concept is not limited to the NR network and can also be applied to other wireless communication systems with similar technical background or channel configuration (for example, cellular communication systems (such as Long Term Evolution (LTE), Advanced LTE (LTE-A), Wireless Broadband (WiBro), and Global System for Mobile Communications (GSM))) and / or short-range communication systems (such as Bluetooth and Near Field Communication (NFC)).

[0023] The wireless communication network of the wireless communication system WCS can support multiple wireless communication devices including the wireless communication device 100 to communicate with each other by sharing available network resources. For example, in the wireless communication network, information can be provided through various multiple access schemes such as code division multiple access (CDMA), frequency division multiple access (FDMA), time division multiple access (TDMA), orthogonal frequency division multiple access (OFDMA), single carrier frequency division multiple access (SC-FDMA), OFDM-FDMA, OFDM-TDMA, OFDM-CDMA.

[0024] In addition, the various functions described below may be implemented or supported by artificial intelligence techniques or one or more computer programs, each of which includes a computer-readable program code and is stored in a non-transitory computer-readable medium. The term "computer-readable medium" includes all types of computer-accessible media (such as read-only memory (ROM), random access memory (RAM), hard disk drives, compact disks (CDs), digital video disks (DVDs), and / or other types of memory). Non-transitory computer-readable media do not include wired communication links, wireless communication links, optical communication links, or other communication links configured to transmit temporary electrical or other signals. Non-transitory computer-readable media include media that can permanently store data, as well as media that store data and can be rewritten later (such as rewritable optical disks or erasable memory devices).

[0025] In the embodiments described below, a hardware method is shown. However, since the embodiments include techniques using both hardware and software, the embodiments do not exclude a software-based method.

[0026] Reference Figure 1 , the wireless communication system WCS may include a unit 200 and / or a wireless communication device 100. The unit 200 may generally indicate a fixed station that communicates with the wireless communication device 100 and / or another unit (not shown), and may exchange data and control information with the wireless communication device 100 and / or another unit (not shown) by communicating with the wireless communication device 100 and / or another unit (not shown). For example, the unit 200 may be referred to as a base station, a node B, an evolved node B (eNB), a next generation node B (gNB), a sector, a site, a base transceiver system (BTS), an access point (AP), a relay node, a remote radio head (RRH), a radio unit (RU), a small cell, a device, etc.

[0027] A user equipment (UE) as an example of the wireless communication device 100 may be fixed or mobile, and may be referred to as any device capable of transmitting data and / or control information to or receiving data and / or control information from the unit 200 by communicating with the unit 200. For example, the UE may be referred to as a terminal, a terminal device, a mobile station (MS), a mobile terminal (MT), a user terminal (UT), a subscriber station (SS), a wireless communication device, a wireless device, a handheld device, etc.

[0028] The unit 200 may provide wireless broadband access to the wireless communication device 100 in the coverage 300 of the unit 200. The wireless communication device 100 may communicate with the unit 200 by generating a transmission signal based on a baseband signal. The baseband signal represents a signal having a relatively low frequency component before modulation. The wireless communication device 100 may generate a transmission signal by amplifying a baseband signal (e.g., a baseband signal modulated to a relatively high frequency) based on a power amplifier. The power amplifier may be a nonlinear element having a nonlinear output characteristic relative to an input. In addition, the wireless communication device 100 may include not only a power amplifier but also at least one nonlinear element (e.g., at least one other nonlinear element). If nonlinearity occurs between the baseband signal and the transmission signal due to the nonlinearity of the wireless communication device 100, the quality of the transmission signal may be degraded. For example, due to the nonlinearity of the wireless communication device 100, the wireless communication device 100 may not be able to send a transmission signal with a desired strength, and because unnecessary (or irrelevant) frequency components are generated due to the intermodulation effect, noise may exist in the transmission signal. Therefore, a method for correcting the nonlinearity of the wireless communication device 100 is needed.

[0029] The wireless communication device 100 according to the inventive concept may use a linear calibration model based on a neural network to correct the nonlinearity of the wireless communication device 100. The wireless communication device 100 according to the inventive concept may generate a calibration signal by performing predistortion, IQ mismatch compensation, and crest factor reduction (CFR) on a baseband signal based on one linear calibration model. The wireless communication device 100 may generate an output signal (or referred to as a transmission signal) based on the calibration signal, thereby correcting the nonlinearity between the baseband signal and the output signal caused by the nonlinear element included in the wireless communication device 100.

[0030] Figure 2 is a block diagram showing a wireless communication device 100 according to an embodiment.

[0031] Figure 2 The wireless communication device 100 is shown with reference to Figure 1 Therefore, reference will not be made here again. Figure 1 Description of the process.

[0032] Reference Figure 2 According to the inventive concept, the wireless communication device 100 may include a processor 110, a transmitter (eg, a radio frequency (RF) transmitter) 120, an antenna 130, and / or a memory 140. Figure 2 The wireless communication device 100 may include various components in addition to the components shown in FIG.

[0033] The wireless communication device 100 may access a wireless communication system by transmitting and receiving signals to and from the wireless communication system via the antenna 130 .

[0034] The processor 110 may control the overall operation of the wireless communication device 100. For example, the processor 110 may be a central processing unit (CPU). The processor 110 may include one processor core (single core) or multiple processor cores (multi-core). The processor 110 may process or execute programs and / or data stored in the memory 140. In an embodiment, the processor 110 may control various functions of the wireless communication device 100 or perform various calculations by executing the programs stored in the memory 140.

[0035] The processor 110 according to the inventive concept may generate a calibration signal by performing pre-distortion and IQ mismatch compensation on the input signal based on a linear calibration model. The linear calibration model may perform pre-distortion and IQ mismatch compensation on the input signal based on a neural network. Specifically, the linear calibration model may include two output layers, and the calibration signal may be generated based on the outputs of the two output layers. Here, the input signal may include a reference Figure 1 The baseband signal described herein is a signal obtained by up-converting a baseband signal from a baseband to a transmission band (radio frequency). Figure 3 The structure of the neural network is described in detail and is referred to below Figure 4 and Figure 5 The structure and operation of the processor 110, predistortion, and IQ mismatch compensation are described in detail.

[0036] The transmitter 120 may be connected to the processor 110. The transmitter 120 may receive a calibration signal from the processor 110. The transmitter 120 may generate an output signal by amplifying the calibration signal. The transmitter 120 may include a power amplifier to be described below, and may amplify the calibration signal based on the power amplifier. In this case, the power amplifier may have nonlinearity, and therefore, as described above, if linear calibration (e.g., predistortion and IQ mismatch compensation) is performed on the input signal based on the processor 110, the nonlinearity of the output signal relative to the input signal may be corrected. Figure 4 A more detailed structure and operation of the transmitter 120 are described. According to an embodiment, the transmitter 120 may be a transceiver.

[0037] The antenna 130 may be connected to the transmitter 120. The antenna 130 may receive the output signal from the transmitter 120. The antenna 130 may transmit the received output signal to another wireless communication device and / or unit 200 (see Figure 1 ). In addition, the antenna 130 may receive a signal from another wireless communication device. Some components for receiving a signal from another wireless communication device through the antenna 130 are Figure 2 In an embodiment, the wireless communication device 100 may include multiple antennas for phased array, multiple input multiple output (MIMO), etc.

[0038] The memory 140 according to the inventive concept may store at least one input signal generated before the current input signal, and send the stored at least one input signal to the processor 110. For example, the memory 140 may store the amplitude of each of the at least one input signal (e.g., the amplitude of the input signal), and output the stored amplitude to the processor 110. The memory 140 may receive the current input signal from the processor 110 and store the received current input signal. For example, the memory 140 may store the amplitude of the current input signal.

[0039] The linear calibration model according to the inventive concept may generate a calibration signal for the current input signal based on the current input signal and at least one previously generated input signal. For example, the processor 110 may receive the amplitude of each of the current input signal and at least one previously generated input signal, and generate a calibration signal for the current input signal based on the linear calibration model.

[0040] Figure 3 is a block diagram illustrating a neural network NN applied to a linear calibration model according to an embodiment.

[0041] The linear calibration model according to the inventive concept is based on a neural network. Figure 3 The neural network NN is used to help understand the neural network structure of the linear calibration model according to the inventive concept, and the inventive concept is not limited thereto.

[0042] Reference Figure 3 The neural network NN may have a structure including an input layer, a hidden layer, and an output layer. The neural network NN may perform calculations based on received input data (e.g., I1 and I2), and generate output data (e.g., O1 and O2) based on the execution results. Here, the input data I1 and I2 may be compared with the reference Figure 2The input signal described corresponds to the input signal described above. For example, the input signal for wireless communication can be represented by a complex number. The input signal represented by a complex number can be called a complex signal. The input data I1 can be the real part of the complex signal, and the input data I2 can be the imaginary part of the complex signal. Here, the real part and the imaginary part of the complex signal can be respectively referred to as in-phase and quadrature phase, and the acronyms based on the in-phase and quadrature phase are called IQ signals. Each of the input signal, calibration signal, and output signal according to the inventive concept can be an IQ signal.

[0043] The neural network NN may be a deep neural network (DNN) or an n-layer neural network including two or more hidden layers. For example, the neural network NN may be a DNN including multiple layers (e.g., an input layer 10, a first hidden layer 12 and a second hidden layer 14, and an output layer 16). However, the present disclosure is not limited thereto. The multiple layers 10, 12, 14, 16 may be implemented by a convolutional layer, a fully connected layer, a softmax layer, etc. For example, a convolutional layer may include convolution, pooling, activation function calculation, etc. Optionally, each of convolution, pooling, and activation function calculation may constitute a layer. However, as described above, the linear calibration model according to the inventive concept is not limited thereto.

[0044] The output of each of the multiple layers 10, 12, 14, and 16 may be referred to as a feature (or feature map). The multiple layers 10, 12, 14, and 16 may receive the features generated in the previous layer as input features, and generate output features or output signals by calculating the input features. Features refer to data representing various characteristics of input data that can be recognized by the neural network NN.

[0045] When the neural network NN has a DNN structure, the neural network NN may include relatively many layers from which effective information is extracted, and thus the neural network NN may process complex data sets. Figure 3 The neural network NN is shown to include four layers (a plurality of layers 10, 12, 14 and 16), but this is merely illustrative and the neural network NN may include fewer or more layers. In addition, the neural network NN may include a plurality of layers having Figure 3 The structures shown in the figure differ in the various layers of the structure.

[0046] Each of the plurality of layers 10, 12, 14, and 16 included in the neural network NN may include a plurality of neurons. Neurons may correspond to artificial nodes referred to as processing elements (PEs), units, or similar terms. For example, Figure 3 As shown in FIG. 1 , the input layer 10 may include two neurons (nodes), and each of the first hidden layer 12 and the second hidden layer 14 may include three neurons (nodes). However, this is merely illustrative, and the layers included in the neural network NN may include a variable number of neurons (nodes), respectively.

[0047] Neurons included in two adjacent layers of the plurality of layers 10, 12, 14, 16 included in the neural network NN may be connected to each other between the two adjacent layers to exchange data. One neuron may receive data from other neurons, calculate the received data, and output the calculation results to other neurons.

[0048] The input and output of each of the neurons (nodes) (e.g., N1, N2, and N3) may be referred to as input activation and output activation, respectively. That is, an activation may be the output of one neuron and, at the same time (or simultaneously), may be a parameter corresponding to the input of a neuron included in a subsequent layer. A neuron may be based on an output activation (e.g., and , weights (e.g., and ) and bias (e.g., and )) to determine the corresponding output activation. Weights and biases are parameters (which may be referred to as calculation parameters) used to calculate the output activation in each neuron, where the weight is the value assigned to the connection relationship between neurons, and the bias represents the weight associated with each neuron. Figure 1 As described, the neural network NN may determine parameters (such as weights and biases) through training (eg, machine learning) based on training data.

[0049] The linear calibration model according to the inventive concept may update weights and biases based on input signals and output signals. Hereinafter, the updating of weights and / or biases (eg, updating of parameters) may be referred to as training the linear calibration model.

[0050] The linear calibration model according to the inventive concept may include two input layers, two output layers and a phase filter. Figure 6 A detailed description of the linear calibration model is given.

[0051] Figure 4 is a block diagram showing a wireless communication device 100 a according to the embodiment.

[0052] Figure 4 The wireless communication device 100a, the processor 110a and the transmitter 120a may be respectively connected to the reference Figure 2 The wireless communication device 100, the processor 110 and the transmitter 120 described above correspond to each other. Therefore, reference will not be repeated here. Figure 2 Description of the process.

[0053] Figure 4The wireless communication device 100a may include a processor 110a and / or a transmitter 120a. The processor 110a may include a linear calibration module 111a, and the transmitter 120a may include a power amplifier 121a and / or a scaling circuit 122a.

[0054] The linear calibration module 111a may include a processing circuit (such as hardware including a logic circuit, a combination of hardware / software (such as a processor for executing software), or a combination thereof). For example, more specifically, the processing circuit may include a CPU, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a microprocessor, an application specific integrated circuit (ASIC), etc., but the inventive concept is not limited thereto.

[0055] The linear calibration module 111a may generate a calibration signal CS by performing linear calibration on the input signal IS according to the linear calibration model 112a based on the neural network. Here, the linear calibration may include CFR, pre-distortion, IQ mismatch compensation, etc. In the inventive concept, the linear calibration may indicate a process for correcting the nonlinearity of the output signal relative to the input signal, which is caused by a nonlinear element included in the wireless communication device 100a. Specifically, the linear calibration may indicate a process for correcting the nonlinearity of the power amplifier 121a.

[0056] The above CFR may indicate a method of correcting nonlinearity due to a peak-to-average power ratio (PAPR) of an input signal. Since the power amplifier 121a has nonlinearity that increases with relatively high power, the nonlinearity of the power amplifier 121a may be corrected by removing a component having relatively high power from among components included in the input signal.

[0057] The above-mentioned pre-distortion is a characteristic opposite to the nonlinearity of the power amplifier 121a, and may indicate a method of pre-distorting the input signal. For example, when the nonlinearity of the power amplifier 121a has a logarithmic function form, the nonlinearity of the power amplifier 121a may be corrected by pre-distorting the input signal into an exponential function form corresponding to an inverse function of the logarithmic function. However, the above-mentioned example is merely illustrative, and the pre-distortion according to the inventive concept is not limited thereto.

[0058] The above-mentioned IQ mismatch compensation may indicate compensation for IQ mismatch. IQ mismatch may include gain mismatch and phase mismatch. The wireless communication device 100a according to the inventive concept may also include a local oscillator (not shown). The local oscillator may be included in the transmitter 120a, but the inventive concept is not limited thereto. The local oscillator may modulate the baseband signal into a carrier frequency signal by up-converting the baseband signal, or modulate the carrier frequency signal into a baseband signal by down-converting the carrier frequency signal. As described above, since the baseband signal includes a real part and an imaginary part, the baseband signal may be referred to as an IQ signal, and when both the real part and the imaginary part, which are orthogonal to each other, are modulated based on the local oscillator, mutual interference may occur due to the error of the local oscillator, resulting in degradation of communication performance. Specifically, gain mismatch may indicate a situation where there is an amplitude difference between the real part and the imaginary part of a signal obtained by modulating a baseband signal by a local oscillator, and phase mismatch may indicate a situation where the real part and the imaginary part are not orthogonal, and due to gain mismatch and phase mismatch, the error vector magnitude (EVM) of the transmission signal may increase, thereby causing communication performance degradation. Therefore, IQ mismatch compensation may be used for transmission performance improvement.

[0059] The linear calibration module 111a according to the inventive concept may generate a calibration signal CS by performing CFR, predistortion and / or IQ mismatch compensation on the input signal IS using a linear calibration model 112a based on a neural network. The linear calibration module 111a may send the calibration signal CS to the transmitter 120a. Figure 6 A detailed description of each of CFR, predistortion, and IQ mismatch compensation based on the linear calibration model 112 a according to the inventive concept is described.

[0060] The transmitter 120a according to the inventive concept may receive a calibration signal CS. The transmitter 120a may generate an output signal OS by amplifying the calibration signal CS based on the amplification factor of the power amplifier 121a. Here, the output signal OS may include an output signal generated by amplifying the input signal IS and an output signal generated by amplifying the calibration signal CS. That is, according to the description, the output signal OS may indicate an output signal generated by amplifying the input signal IS or an output signal generated by amplifying the calibration signal CS. In order to distinguish from the output signal generated by amplifying the input signal IS, the output signal generated by amplifying the calibration signal CS may be referred to as a calibrated output signal. In this case, the power amplifier 121a has the nonlinearity as described above, and the calibration signal CS is a signal that has been subjected to linear calibration including CFR, pre-distortion, and IQ mismatch compensation, so the output signal OS generated by amplifying the calibration signal CS may be a signal obtained by correcting the nonlinearity of the power amplifier 121a (therefore, the output signal OS may be referred to as a calibrated output signal). The calibrated output signal may represent the above-mentioned transmission signal. Therefore, as described above with reference to Figure 2 As described, the calibrated output signal can be transmitted through antenna 130 (refer to Figure 2 ). According to an embodiment, the wireless communication device 100a may send the calibrated output signal to the unit 200 and / or another wireless communication device 100a. According to an embodiment, in response to the sent calibrated output signal, the wireless communication device 100a may receive a reply signal (e.g., from the unit 200 and / or another wireless communication device 100a). The wireless communication device 100a may process the reply signal (e.g., down-convert, demodulate, etc.) to obtain a message contained in the reply signal.

[0061] The power amplifier 121a may send the output signal OS to the scaling circuit 122a. The scaling circuit 122a may output a scaled output signal SS (or a reduced output signal) by scaling the output signal OS by the inverse of the amplification factor of the power amplifier 121a. The scaled output signal SS may have the same amplitude (or a similar amplitude) as the input signal IS. For example, according to the embodiments described below, the scaling circuit 122a may generate the scaled output signal SS by scaling at least one of the output signal based on the input signal IS and the output signal based on the calibration signal CS.

[0062] The wireless communication device 100a according to the inventive concept may also include a local oscillator as described above, and the local oscillator may down-convert the scaled output signal SS. The frequency band of the scaled output signal SS that has been down-converted may be the same (or similar) as the frequency band of the input signal IS. Hereinafter, the wireless communication device 100a will be described to train the linear calibration model 112a based on the input signal IS and the output signal OS (e.g., the calibrated output signal). Here, for ease of description, it is assumed that the output signal OS indicates the scaled output signal SS, and the frequency band of the input signal IS is the same (or similar) as the frequency band of the output signal OS. However, this is for ease of description, and the inventive concept is not limited thereto.

[0063] The linear calibration module 111 a may train a linear calibration model 112 a based on the input signal IS and the scaled output signal SS.

[0064] The wireless communication device 100a according to the inventive concept can correct the nonlinearity of the wireless communication device 100a by performing linear calibration on the input signal (e.g., the input signal generated after the input signal IS as described above) based on the trained linear calibration model 112a. That is, the wireless communication device 100a can train the linear calibration model 112a based on the input signal IS and the output signal OS generated by the power amplifier 121a, and as described above, can repeat the training based on a plurality of input signals to improve the nonlinear correction performance of the wireless communication device 100a using the linear calibration model 112a.

[0065] Figure 5 is a block diagram showing a wireless communication device 100b according to the embodiment.

[0066] Figure 5 The wireless communication device 100b may include a processor 110b and / or a transmitter 120b. The processor 110b may include a linear calibration module 111b and / or a power amplifier estimation module 113b, and the transmitter 120b may include a power amplifier 121b and / or a scaling circuit 122b.

[0067] because Figure 5 The wireless communication device 100b, the processor 110b, the transmitter 120b, the linear calibration module 111b, the power amplifier 121b and the scaling circuit 122b are respectively Figure 4 The wireless communication device 100a, the processor 110a, the transmitter 120a, the linear calibration module 111a, the power amplifier 121a and the scaling circuit 122a correspond to each other, so repeated descriptions are omitted here.

[0068] Reference Figure 4 and Figure 5 , Figure 5The wireless communication device 100b may further include a power amplifier estimation module 113b.

[0069] The power amplifier estimation module 113b according to the inventive concept may generate an estimated output signal EOS based on a power amplifier estimation model and a calibration signal CS. The power amplifier estimation model may be based on a neural network (refer to Figure 3 ). However, the inventive concept is not limited thereto, and the power amplifier estimation model may be based on a memory polynomial (MP) or a generalized MP (GMP). The power amplifier estimation module 113b according to the inventive concept may train the power amplifier estimation model based on the input signal IS and the output signal OS. Here, the output signal OS may be a signal obtained by amplifying the input signal IS based on the amplification factor. However, the inventive concept is not limited thereto. The linear calibration module 111b according to the inventive concept may train the linear calibration model 112b based on the estimated output signal EOS and the input signal IS.

[0070] The power amplifier estimation module 113b according to the inventive concept may generate an estimated output signal EOS based on a trained power amplifier estimation model and a calibration signal CS. The estimated output signal EOS may be similar to or identical to an output signal OS (e.g., a calibrated output signal) generated by the power amplifier 121b based on the calibration signal CS. As described above, the power amplifier estimation module 113b may train a power amplifier estimation model (e.g., update parameters (such as weights and biases) of the power amplifier estimation model) based on an input signal IS and an output signal obtained by amplifying the input signal IS based on an amplification factor, and may repeatedly train the power amplifier estimation model by repeating this operation for each of a plurality of input signals. A fully trained power amplifier estimation model may operate similarly or identically to the power amplifier 121b.

[0071] Therefore, even when the output signal OS is not generated by the power amplifier 121b, the wireless communication device 100b according to the inventive concept can train the linear calibration model 112b based on the input signal IS and the estimated output signal EOS "generated based on the trained power amplifier estimation model". The wireless communication device 100b can generate the output signal OS calibrated based on the trained linear calibration model 112b, and the calibrated output signal OS can be a signal that has been corrected for nonlinearity by the wireless communication device 100b. According to an embodiment, the wireless communication device 100b can send the output signal OS to the unit 200 and / or another wireless communication device 100b. According to an embodiment, in response to the transmitted output signal OS, the wireless communication device 100b can receive a response signal (e.g., from the unit 200 and / or another wireless communication device 100b). The wireless communication device 100b can process the response signal (e.g., down-convert, demodulate, etc.) to obtain a message contained in the response signal.

[0072] Figure 6 is a block diagram illustrating a linear calibration model 112c according to an embodiment.

[0073] Figure 6 The structure of the linear calibration model 112c shown in FIG. 1 (eg, the hidden layer HL, the first output layer OL_1, the second output layer OL_2, and the nodes in each layer) can be obtained by referring to Figure 3 The neural network NN described is easy to understand.

[0074] Although not shown for ease of illustration, reference Figure 3 Understandable, Figure 6 The arrows between the input data (e.g., Re[IS(n)], Im[IS(n)], ABS[IS(n)], etc.) and the nodes in the figure may indicate that weights are applied to the input data, and the outputs of the nodes may indicate that biases are applied to the input data. In addition, it can be understood that the linear calibration model 112c may include an input layer and at least one hidden layer HL.

[0075] The linear calibration model 112c can be used with Figure 4 The linear calibration model 112a and / or Figure 5 The linear calibration model 112b corresponds to .

[0076] Reference Figure 6 , the linear calibration model 112c according to the inventive concept may generate a calibration signal based on the current input signal IS(n). The current input signal IS(n) and the calibration signal may be complex signals. For example, the first calibration value Re[CS] may be the real part of the calibration signal, and the second calibration value Im[CS] may be the imaginary part of the calibration signal.

[0077] The linear calibration model 112c may generate a first IQ compensation value IQC_1 and a second IQ compensation value IQC_2 based on the current input signal IS(n). Specifically, the linear calibration model 112c may include a first output layer OL_1, and may generate a first IQ compensation value IQC_1 and a second IQ compensation value IQC_2 based on a real part Re[IS(n)] of the current input signal IS(n) input through the first output layer OL_1 and an imaginary part Im[IS(n)] of the current input signal IS(n). The first IQ compensation value IQC_1 and the second IQ compensation value IQC_2 may indicate values ​​obtained by performing the above-mentioned IQ mismatch compensation on the current input signal IS(n).

[0078] The linear calibration model 112c may generate a first predistortion value PD_1 and a second predistortion value PD_2 based on a current input signal IS(n) and at least one previously generated input signal. Specifically, the linear calibration model 112c may include at least one hidden layer HL and a second output layer OL_2, and may generate a plurality of output values ​​OV_1 to OV_2M based on respective amplitudes ABS[IS(n)] to ABS[IS(n-M+1)] of a plurality of input signals input through the at least one hidden layer HL and the second output layer OL_2. The second output layer OL_2 may send the plurality of output values ​​OV_1 to OV_2M to the phase filter PF. Although the number of output values ​​OV_1 to OV_2M is represented as twice the number of corresponding amplitudes ABS[IS(n)] to ABS[IS(n-M+1)] of the plurality of input signals, this is merely illustrative and the inventive concept is not limited thereto. The output values ​​OV_1 to OV_2M may be retrieved from the memory 140 (refer to Figure 2 ) receives the corresponding amplitudes ABS[IS(n-1)] to ABS[IS(n-M+1)] of the previously generated input signals. Here, n can be used to indicate the current time point. Therefore, the current input signal IS(n) can indicate the most recently generated input signal. M is an integer greater than or equal to 2.

[0079] Power amplifier 121a (see Figure 4 ) may be affected by previously generated input signals. Therefore, the linear calibration model 112c can effectively correct the nonlinearity caused by the power amplifier 121a (refer to Figure 4 ) caused by nonlinearity.

[0080] The phase filter PF may generate a first predistortion value PD_1 and a second predistortion value PD_2 based on a plurality of output values ​​OV_1 to OV_2M and a plurality of input signals IS(n) to IS(n-M+1). For ease of description, previously generated input signals IS(n-1) to IS(n-M+1) received by the phase filter PF are omitted. The first predistortion value PD_1 and the second predistortion value PD_2 may indicate values ​​obtained by performing predistortion on the current input signal IS(n).

[0081] The first adder 410 and the second adder 420 may generate a first calibration value Re[CS] and a second calibration value Im[CS] based on the first IQ compensation value IQC_1, the second IQ compensation value IQC_2, the first predistortion value PD_1, and the second predistortion value PD_2, respectively. Specifically, the first adder 410 may generate the first calibration value Re[CS] by adding the first IQ compensation value IQC_1 to the first predistortion value PD_1. The second adder 420 may generate the second calibration value Im[CS] by adding the second IQ compensation value IQC_2 to the second predistortion value PD_2.

[0082] As described above, the linear calibration model 112c according to the inventive concept may generate a calibration signal based on the first calibration value Re[CS] and the second calibration value Im[CS] (eg, by combining the first calibration value Re[CS] and the second calibration value Im[CS]), and the power amplifier 121a (refer to Figure 4 ) can generate an output signal based on the calibration signal. The linear calibration model 112c according to the inventive concept can update the parameters through the loss function based on the current input signal IS(n) and the output signal. The processor 110a according to the inventive concept (refer to Figure 4 ) may calculate a loss function as represented by Equation 1. However, the inventive concept is not limited thereto.

[0083] [Equation 1]

[0084] As described above, the scaling circuit 122a (see Figure 4 ) can generate a scaled output signal SS by scaling down the output signal based on the inverse of the amplification factor (refer to Figure 4 ).

[0085] In Equation 1, the mean square error (MSE) indicates the mean square of the difference between the input signal and the scaled output signal. Here, the frequency band of the input signal may be the same as (or similar to) the frequency band of the scaled output signal. The difference between the input signal and the scaled output signal may be calculated by a loss function, and the parameters of the linear calibration model 112c may be updated so that the difference is minimized (or reduced). Here, the parameters may be updated while minimizing (or reducing) the difference so that the above-mentioned CFR is performed. That is, the PAPR may be reduced by performing CFR on the input signal through the linear calibration model 112c that has been trained.

[0086] In Equation 1, the adjacent channel leakage ratio (ACLR) is an indicator indicating linearity, and indicates a ratio of adjacent channel power of the output signal to the channel power. As described above, if intermodulation occurs due to the nonlinearity of the power amplifier, an unexpected frequency component (which may be referred to as noise) may be generated in a channel adjacent to the channel of the output signal. The processor 110a (refer to Figure 4 ) can calculate ACLR based on the output signal.

[0087] In Equation 1, EVM is an index indicating linearity and is obtained by measuring the difference between an ideal modulation value of an input signal and an actual modulation value of the input signal. Therefore, due to the above-mentioned IQ mismatch and nonlinearity of the power amplifier, the EVM may be worse. The processor 110a (refer to Figure 4 ) can calculate the EVM based on the output signal and the input signal. The processor 110a (refer to Figure 4 ) can update the parameters of the linear calibration model 112a based on the loss function so that MSE, EVM and ACLR are minimized (or reduced). The processor 110a (refer to Figure 4 ) can calculate the loss function based on the above indicators and other indicators. For example, processor 110a (refer to Figure 4 ) The loss function can be calculated by adding the sensitivity of the receiver.

[0088] In equation 1, and are the respective balancing factors of ACLR and EVM. The processor 110a (refer to Figure 4 ) can adjust the respective balance factors of ACLR and EVM according to the characteristics and / or specifications of the signal. The processor 110a (refer to Figure 4 ) can complementarily adjust the balance factors to be applied to ACLR and EVM respectively. By processor 110a (refer to Figure 4) Complementarily adjusting the balancing factor may indicate that when the balancing factor to be applied to either EVM and / or ACLR is reduced or increased, the balancing factor to be applied to the other of EVM and / or ACLR is increased or decreased. For example, referring to Equation 1, if EVM is small enough when ACLR is further reduced to meet a specific specification, then Can be enlarged and According to an embodiment, the processor 110a may update the parameters of the linear calibration model by complementarily adjusting the balance factor (eg, based on a loss function in which the balance factor has been adjusted). According to an embodiment, the balance factor may be omitted from Equation 1, or both may be set equal to "1".

[0089] Figure 7 is a graph comparing the ACLR of an output signal generated by the wireless communication device according to the embodiment with the ACLR of an output signal according to a comparative example.

[0090] Figure 7 is a graph comparing the ACLR of an output signal generated by a neural network-based linear calibration model according to the inventive concept with the ACLR of an output signal according to a comparative example.

[0091] exist Figure 7 , the horizontal axis represents the frequency band and the vertical axis represents the power spectrum. Since the power spectrum is smaller in the frequency band (surrounding frequency band) other than the specific frequency band (which may be referred to as the target frequency band), the ACLR may be smaller. Specifically, the ACLR may decrease as the power spectrum of the surrounding frequency band relatively close to the target frequency band among the surrounding frequency bands decreases.

[0092] Reference Figure 7 Among the ACLR 520 of the output signal generated by the linear calibration model based on the neural network according to the inventive concept, the ACLR 510 of the output signal without pre-distortion, the ACLR 540 of the output signal in which pre-distortion is performed on the output signal by the MP model, and the ACLR 530 of the output signal in which pre-distortion is performed on the output signal by the memoryless polynomial model, the ACLR 520 of the output signal generated by the linear calibration model based on the neural network according to the inventive concept is the smallest.

[0093] Figure 8 is a flowchart for describing an operation method of a wireless communication device according to an embodiment.

[0094] Reference Figure 8 In operation S100, the wireless communication device 100 (refer to Figure 1) An IQ compensation value may be generated by performing IQ mismatch compensation on a first input signal using a linear calibration model based on a neural network. The IQ compensation value may include a first IQ compensation value corresponding to a real part of the calibration signal and a second IQ compensation value corresponding to an imaginary part of the calibration signal, the first IQ compensation value and the second IQ compensation value having been referenced to Figure 6 Described.

[0095] In operation S200, the wireless communication device 100 (refer to Figure 1 ) can generate a predistortion value by performing predistortion on the first input signal using a linear calibration model. The predistortion value may include a first predistortion value corresponding to the real part of the calibration signal and a second predistortion value corresponding to the imaginary part of the calibration signal, the first predistortion value and the second predistortion value having been referenced to Figure 6 Described.

[0096] In operation S300, the wireless communication device 100 (refer to Figure 1 ) can generate a first calibration signal based on the IQ compensation value and the predistortion value. The wireless communication device 100 (refer to Figure 1 ) The real part of the first calibration signal may be calculated based on the sum of the first IQ compensation value and the first predistortion value, and the imaginary part of the first calibration signal may be calculated based on the sum of the second IQ compensation value and the second predistortion value. As described above, each of the input signal, the calibration signal, and the output signal may be a complex signal.

[0097] In operation S400, the wireless communication device 100 (refer to Figure 1 ) can generate a calibrated first output signal by amplifying the first calibration signal based on the amplification factor.

[0098] In operation S500, the wireless communication device 100 (refer to Figure 1 ) can train a linear calibration model based on the first input signal and the calibrated first output signal. That is, the wireless communication device 100 (refer to Figure 1 ) can update parameters of the linear calibration model based on the first input signal and the calibrated first output signal.

[0099] Wireless communication device 100 (see Figure 1 ) can perform the above-mentioned IQ mismatch compensation and pre-distortion on the input signal based on the trained linear calibration model. For example, as described above, the wireless communication device 100 (refer to Figure 1) can update parameters for linear calibration based on the first input signal and the calibrated first output signal, and generate a second calibration signal by performing IQ mismatch compensation and predistortion on a second input signal generated after the first input signal based on the trained linear calibration model. The power amplifier can generate a second output signal by amplifying the second calibration signal. Since the second calibration signal is a signal pre-calibrated corresponding to the nonlinearity of the power amplifier, the nonlinearity of the second output signal relative to the second input signal can be corrected regardless of the nonlinearity of the power amplifier.

[0100] As described above, in operation S500, the wireless communication device 100 (refer to Figure 1 ) can train the linear calibration model so that the difference between the input signal and the output signal corresponding to the input signal is minimized (or reduced). That is, the parameters of the linear calibration model can be updated. In addition, the EVM can be calculated based on the input signal and the calibrated output signal (for example, by the processor 110), and the ACLR can be calculated based on the calibrated output signal (for example, by the processor 110). The wireless communication device 100 (refer to Figure 1 ) can train a linear calibration model so that EVM and ACLR are minimized (or reduced), and also by (for example, using a combination of Figure 6 The loss function discussed above) applies a balancing factor to each of EVM and ACLR while training a linear calibration model based on the specific specifications and / or characteristics of the signal.

[0101] Fig. 9 is a flowchart for describing an operation method of a wireless communication device according to an embodiment.

[0102] Fig. 9 The description of operations S100, S200, S300 and S400 is the same as that of the above reference Figure 8 The descriptions are the same (or similar) and thus are omitted here.

[0103] However, the operations described below may be performed without performing operations S100, S200, S300, and S400. That is, operation S600 may be performed separately from operations S100, S200, S300, and S400.

[0104] Reference Fig. 9 In operation S600, a power amplifier estimation model based on a neural network may be trained based on a first input signal and an output signal obtained by amplifying the first input signal based on an amplification factor. The trained power amplifier estimation model may operate similarly or identically to the power amplifier. This is an example according to the inventive concept, and the inventive concept is not limited thereto. For example, the power amplifier estimation model may be trained based on the first input signal and the calibrated first output signal.

[0105] The wireless communication device 100 (see Figure 1 ) can be performed by performing a reference on a second input signal generated after the first input signal based on a linear calibration model Figure 8 The wireless communication device 100 (refer to Figure 1 ) can generate a first estimated output signal by amplifying the second calibration signal based on the trained power amplifier estimation model. The wireless communication device 100 (refer to Figure 1 ) can train a linear calibration model based on the first estimated output signal and the second input signal.

[0106] The wireless communication device 100 (see Figure 1 ) A third calibration signal can be generated by performing IQ mismatch compensation and predistortion on a third input signal generated after the second input signal based on a trained linear calibration model, and a third output signal can be generated by amplifying the third calibration signal. Since the third calibration signal is a signal pre-calibrated corresponding to the nonlinearity of the power amplifier, the nonlinearity of the third output signal relative to the third input signal can be corrected regardless of the nonlinearity of the power amplifier.

[0107] The wireless communication device 100 (see Figure 1 ) can train the linear calibration model so that the difference between the second input signal and the first estimated output signal is minimized (or reduced). As described above, the wireless communication device 100 (refer to Figure 1 ) can update the parameters of the linear calibration model so that the MSE of the second input signal and the first estimated output signal is minimized (or reduced). In addition, the wireless communication device 100 (refer to Figure 1 ) can calculate the EVM based on the second input signal and the first estimated output signal, and calculate the ACLR based on the first estimated output signal. The wireless communication device 100 (refer to Figure 1 ) can train a linear calibration model based on any one of the calculated EVM and / or ACLR. For example, the wireless communication device 100 (refer to Figure 1 ) can update the parameters of the linear calibration model so that EVM and ACLR are minimized (or reduced). In addition, the wireless communication device 100 (refer to Figure 1 ) A balancing factor may be applied to each of EVM and ACLR, and a linear calibration model that considers the characteristics of the signal and the required (or configured) specifications may be trained by adjusting the balancing factor applied to each of EVM and ACLR.

[0108] In the description made above, each of “first”, “second”, and “third” is an expression for identifying a signal, and does not limit the number of specific signals in the inventive concept.

[0109] Fig.10 is a block diagram showing a wireless communication device 1000 according to an embodiment.

[0110] Reference Fig.10 , the wireless communication device 1000 may include an ASIC 1100, an application specific instruction set processor (ASIC) 1300, a memory 1500, a main processor 1700, and / or a main memory 1900. Two or more of the ASIC 1100, the ASIP 1300, and the main processor 1700 may communicate with each other. In addition, two or more of the ASIC 1100, the ASIP 1300, the memory 1500, the main processor 1700, and the main memory 1900 may be embedded in one chip.

[0111] ASIP 1300, which is an integrated circuit customized for a specific purpose, may support a dedicated instruction set for a specific application and execute instructions included in the instruction set. Memory 1500 may communicate with ASIP 1300 and store a plurality of instructions to be executed by ASIP 1300 as a non-transitory storage device. For example, as a non-limiting example, memory 1500 may include a random type of memory (such as RAM, ROM, tape, disk, optical disk, volatile memory, non-volatile memory, or a combination thereof) that can be accessed by ASIP 1300.

[0112] The main processor 1700 may control the wireless communication device 1000 by executing a plurality of instructions. For example, the main processor 1700 may control the ASIC 1100 and the ASIP 1300, and process received data or user input on the wireless communication device 1000. The main memory 1900 may communicate with the main processor 1700, and store a plurality of instructions to be executed by the main processor 1700 as a non-transitory storage device. For example, as a non-limiting example, the main memory 1900 may include a random type of memory such as RAM, ROM, tape, disk, optical disk, volatile memory, non-volatile memory, or a combination thereof that can be accessed by the main processor 1700.

[0113] Already referred to Figures 1 to 9 The described “wireless communication device and operation method of the wireless communication device according to the embodiment” can be included in Fig.10In an embodiment, at least one operation in the above-mentioned operation method of the wireless communication device 1000 may be implemented by a plurality of instructions stored in the memory 1500. In an embodiment, at least one operation in the operation method may be performed by the ASIP 1300 executing a plurality of instructions stored in the memory 1500.

[0114] A wireless communication device generates a transmission signal by amplifying a baseband signal using a power amplifier. However, the power amplifier has a nonlinear output characteristic relative to the input signal, so that the generated transmission signal has reduced signal strength and excessive noise. Conventional devices and methods for correcting the nonlinearity of the power amplifier rely on inaccurate heuristic methods that cannot fully compensate for the nonlinearity. Therefore, conventional devices and methods suffer from excessive performance degradation including, for example, transmission signals with reduced signal strength and excessive noise.

[0115] However, according to an embodiment, an improved device and method are provided for correcting the nonlinearity of a power amplifier. For example, the improved device and method may train a linear calibration model based on an input signal to output a calibration signal so that when the calibration signal is amplified by the power amplifier, the calibration signal corresponds to the input signal amplified when the nonlinearity of the power amplifier is corrected. For example, the linear calibration model may be trained to minimize (or reduce) the difference between the input signal and the amplified calibration signal. Therefore, the improved device and method may overcome the defects of conventional devices and methods to at least reduce the performance degradation caused by the nonlinearity of the power amplifier. In doing so, the improved device and method may increase the signal strength of the transmission signal generated by the improved device and method, and reduce the noise included in the transmission signal generated by the improved device and method, relative to the transmission signal generated by the conventional device and method.

[0116] According to an embodiment, the operations described herein as being performed by the wireless communication system WCS, the unit 200, the wireless communication device 100, the processor 110, the transmitter 120, the wireless communication device 100a, the processor 110a, the transmitter 120a, the linear calibration module 111a, the scaling circuit 122a, the wireless communication device 100b, the processor 110b, the transmitter 120b, the linear calibration module 111b, the power amplifier estimation module 113b, the scaling circuit 122b, the phase filter PF, the first adder 410, the second adder 420, the wireless communication device 1000, the ASIC 1100, the ASIP 1300 and / or the main processor 1700 may be performed by a processing circuit. In addition to the above description, the term "processing circuit" used in the present disclosure may refer to, for example, hardware including a logic circuit, a hardware / software combination (such as a processor executing software), or a combination thereof. For example, the processing circuit may more specifically include, but is not limited to, a central processing unit (CPU), an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a system on a chip (SoC), a programmable logic unit, a microprocessor, an application specific integrated circuit (ASIC), etc.

[0117] In an embodiment, the processing circuit may perform some operations (e.g., operations described herein as being performed by the linear calibration model 112a, the linear calibration model 112b, the power amplifier estimation model, the linear calibration model 112c, the output layer OL_1, the hidden layer HL, and / or the output layer OL_2) through an artificial intelligence model and / or a machine learning model. As an example, the processing circuit may implement an artificial neural network (e.g., the linear calibration model 112a, the linear calibration model 112b, the power amplifier estimation model, and / or the linear calibration model 112c) trained on a set of training data through, for example, a supervised, unsupervised, and / or reinforcement learning model, and the processing circuit may process the feature vector to provide an output based on the training. Such an artificial neural network may utilize various artificial neural network organization and processing models (such as, convolutional neural network (CNN), recurrent neural network (RNN) optionally including long short-term memory (LSTM) units and / or gated recurrent units (GRU), stacked deep neural network (S-DNN), state-space dynamic neural network (S-SDNN), deconvolution network, deep belief network (DBN) and / or restricted Boltzmann machine (RBM)). Alternatively or additionally, the processing circuit may include other forms of artificial intelligence models and / or machine learning models (such as, for example, linear and / or logistic regression, statistical clustering, Bayesian classification, decision trees, dimensionality reduction (such as principal component analysis) and expert systems; and / or combinations thereof (including ensembles such as random forests)).

[0118] Here, the machine learning model may have any structure that is trainable (e.g., using training data). For example, the machine learning model may include an artificial neural network, a decision tree, a support vector machine, a Bayesian network, a genetic algorithm, etc. The machine learning model will now be described primarily with reference to an artificial neural network, but the embodiments are not limited thereto. Non-limiting examples of artificial neural networks may include convolutional neural networks (CNNs), region-based convolutional neural networks (R-CNNs), region proposal networks (RPNs), recursive neural networks (RNNs), stacked deep neural networks (S-DNNs), state-space dynamic neural networks (S-SDNNs), deconvolution networks, deep belief networks (DBNs), restricted Boltzmann machines (RBMs), fully convolutional networks, long short-term memory (LSTM) networks, classification networks, etc.

[0119] The various operations of the above methods may be performed by any suitable device capable of performing the operations (such as the processing circuits discussed above). For example, as described above, the operations of the above methods may be performed by various hardware and / or software implemented in some form of hardware (e.g., processor, ASIC, etc.).

[0120] The software may include an ordered listing of executable instructions for implementing logical functions and may be implemented in any "processor readable medium" for use by or in conjunction with an instruction execution system, device or apparatus (such as a single-core processor or multi-core processor or a system including a processor).

[0121] The blocks or operations of the methods or algorithms and functions described in conjunction with the embodiments disclosed herein may be implemented directly in hardware, in software modules executed by a processor, or in a combination of the two. If implemented in software, the functions may be stored as one or more instructions or codes on or transmitted through a tangible, non-transitory computer-readable medium (e.g., memory 140, memory 1500, and / or main memory 1900). The software module may reside in a random access memory (RAM), flash memory, read-only memory (ROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), registers, hard disk, removable disk, CD ROM, or any other form of storage medium known in the art.

[0122] While the inventive concept has been particularly shown and described with reference to embodiments thereof, it will be understood that various changes in form and details may be made therein without departing from the spirit and scope of the appended claims.

Claims

1. A wireless communication device, comprising: a first processing circuit configured to generate a calibration signal by performing a linear calibration on the first input signal using a linear calibration model, the linear calibration model being based on a first neural network; as well as a power amplifier configured to generate a calibrated output signal by amplifying the calibration signal based on an amplification factor, in, The linear calibration model includes: a first output layer; a first output layer configured to generate a first IQ compensation value and a second IQ compensation value by performing IQ mismatch compensation on the first input signal, the first IQ compensation value corresponding to a real part of the calibrated output signal, and the second IQ compensation value corresponding to an imaginary part of the calibrated output signal, and a second processing circuit configured to generate a first predistortion value and a second predistortion value by performing predistortion on the first input signal, the first predistortion value corresponding to a real part of the calibrated output signal and the second predistortion value corresponding to an imaginary part of the calibrated output signal, and The first processing circuit is configured to generate a calibration signal based on the first IQ compensation value, the second IQ compensation value, the first pre-distortion value, and the second pre-distortion value.

2. The wireless communication device according to claim 1, wherein: The first processing circuit is configured to update parameters of the linear calibration model based on the first input signal and the calibrated output signal.

3. The wireless communication device according to claim 2, wherein: The first processing circuit is configured to: generating a scaled output signal by scaling down the calibrated output signal based on the inverse of the amplification factor; and Parameters of the linear calibration model are updated based on a difference between the scaled output signal and the first input signal.

4. The wireless communication device according to claim 3, wherein: The first processing circuit is configured to: calculating an error vector magnitude based on the first input signal and the calibrated output signal; calculating an adjacent channel leakage ratio based on the calibrated output signal; and Parameters of the linear calibration model are updated based on at least one of an error vector magnitude or an adjacent channel leakage ratio.

5. The wireless communication device according to claim 4, wherein: The first processing circuit is configured to: applying a first balancing factor to the error vector magnitude and a second balancing factor to the adjacent channel leakage ratio; and The parameters of the linear calibration model are updated by complementarily adjusting the first balance factor and the second balance factor.

6. The wireless communication device according to claim 1, further comprising: Power amplifier estimation model based on the second neural network, The first processing circuit is configured to update parameters of the power amplifier estimation model based on a first input signal and a first output signal, and the first output signal is obtained by amplifying the first input signal based on an amplification factor.

7. The wireless communication device according to claim 6, wherein: The power amplifier estimation model is configured to generate an estimated output signal based on the calibration signal; as well as The first processing circuit is configured to update parameters of the linear calibration model based on the first input signal and the estimated output signal.

8. The wireless communication device according to claim 6, wherein: The first processing circuit is configured to: calculating an error vector magnitude based on the first input signal and the estimated output signal; calculating an adjacent channel leakage ratio based on the estimated output signal; as well as Parameters of the linear calibration model are updated based on at least one of an error vector magnitude or an adjacent channel leakage ratio.

9. The wireless communication device according to claim 1, wherein: The real part of the calibration signal is the sum of the first IQ compensation value and the first predistortion value; and The imaginary part of the calibration signal is the sum of the second IQ compensation value and the second predistortion value.

10. The wireless communication device according to claim 1, wherein: The linear calibration model further includes a second output layer configured to output at least two output values, the at least two output values ​​being generated based on an amplitude of the first input signal and an amplitude of each of at least one second input signal, and the at least one second input signal being generated before the first input signal; and The second processing circuit is configured to generate a first predistortion value and a second predistortion value based on the first input signal and the at least two output values.

11. A method for operating a wireless communication device, the method comprising: generating an IQ compensation value by performing IQ mismatch compensation on a first input signal using a linear calibration model, the linear calibration model being based on a neural network; generating predistortion values ​​by performing predistortion on the first input signal using a linear calibration model; generating a first calibration signal based on the IQ compensation value and the predistortion value; generating a calibrated first output signal by amplifying the first calibration signal based on the amplification factor; as well as A linear calibration model is trained based on the first input signal and the calibrated first output signal.

12. The method according to claim 11, wherein: The IQ compensation value includes a first IQ compensation value and a second IQ compensation value, the first IQ compensation value corresponds to a real part of the calibrated first output signal, and the second IQ compensation value corresponds to an imaginary part of the calibrated first output signal; The predistortion value comprises a first predistortion value and a second predistortion value, the first predistortion value corresponds to a real part of the calibrated first output signal, and the second predistortion value corresponds to an imaginary part of the calibrated first output signal; as well as The step of generating a first calibration signal comprises generating the first calibration signal based on: The sum of the first IQ compensation value and the first predistortion value, and The sum of the second IQ compensation value and the second predistortion value.

13. The method according to claim 11, further comprising: After training of the linear calibration model, IQ mismatch compensation and predistortion are performed on a second input signal, the second input signal being generated after the first input signal.

14. The method according to claim 11, wherein: The steps to train a linear calibration model include: training a linear calibration model such that a difference between the first input signal and the calibrated first output signal is minimized; and The linear calibration model is also trained based on at least one of an error vector magnitude calculated based on the first input signal and the calibrated first output signal and an adjacent channel leakage ratio calculated based on the calibrated first output signal.

15. The method according to claim 14, wherein: The steps to train a linear calibration model include: applying a first balancing factor to the error vector magnitude and a second balancing factor to the adjacent channel leakage ratio; and The linear calibration model is trained by complementarily adjusting the first balance factor and the second balance factor.

16. A method for operating a wireless communication device, the method comprising: generating an IQ compensation value by performing IQ mismatch compensation on a first input signal using a linear calibration model, the linear calibration model being based on a first neural network; generating predistortion values ​​by performing predistortion on the first input signal using a linear calibration model; generating a first calibration signal based on the IQ compensation value and the predistortion value; generating a calibrated first output signal by amplifying the first calibration signal based on the amplification factor; as well as A power amplifier estimation model is trained based on the first input signal and the first output signal to obtain a trained power amplifier estimation model, the power amplifier estimation model is based on a second neural network, and the first output signal is obtained by amplifying the first input signal based on an amplification factor.

17. The method according to claim 16, further comprising: generating a second calibration signal by performing IQ mismatch compensation and predistortion on a second input signal using a linear calibration model, the second input signal being generated after the first input signal; as well as A first estimated output signal is generated by amplifying the second calibration signal using the trained power amplifier estimation model.

18. The method according to claim 17, further comprising: A linear calibration model is trained based on the first estimated output signal and the second input signal.

19. The method according to claim 18, further comprising: generating a third calibration signal by performing IQ mismatch compensation and predistortion on a third input signal using the linear calibration model after training the linear calibration model, the third input signal being generated after the second input signal; as well as A third output signal is generated by amplifying the third calibration signal based on the amplification factor.

20. The method according to claim 18, wherein: The steps to train a linear calibration model include: training a linear calibration model to minimize a difference between the second input signal and the first estimated output signal; and The linear calibration model is also trained based on at least one of an error vector magnitude calculated based on the second input signal and the first estimated output signal and an adjacent channel leakage ratio calculated based on the first estimated output signal.

Images