Multi-channel transmitter, communication equipment and distortion compensation method

Abstract

The invention provides a multi-channel transmitter which comprises a processing module, a first pre-distortion module comprising a first pre-distortion model and a second pre-distortion module comprising a second pre-distortion model, and the second pre-distortion model is generated according to training of the first pre-distortion model. The first pre-distortion model performs first pre-distortion processing on the input signals of the plurality of channels to obtain a first pre-distortion signal of a specified channel and first pre-distortion signals of other channels; the second pre-distortion model at least performs second pre-distortion processing on the input signals of the plurality of channels to obtain second pre-distortion signals of other channels; the processing module processes the second pre-distortion signal of each other channel and the first pre-distortion signal of each other channel to obtain a third pre-distortion signal of each other channel; the pre-distorter structure of a multi-channel transmitter can be simplified, computing power resources in a multi-channel scene are saved, and the precision of pre-distortion compensation is improved. The invention also provides communication equipment and a distortion compensation method.

Description

Technical Field

[0001] The present disclosure relates to the field of communication technology, and in particular to a multi-channel transmitter, communication equipment and a distortion compensation method. Background Art

[0002] In the transmitter, the power amplifier (PA) is one of the devices that consumes the most power. In order to save costs, the working efficiency of the PA should be improved as much as possible in practical applications. At this time, the linearity of the PA will deteriorate, which will aggravate the signal distortion. In order to take into account both the efficiency and linearity of the PA, the digital predistortion (DPD) technology adds an inverse model of the PA behavior to the front end of the PA. Before the signal is input into the PA, the signal is distorted in advance to offset the nonlinear distortion of the PA. With the continuous development of communication systems, large-scale MIMO (Multiple-inmultiple-out) technology has gradually matured. The application of traditional digital predistortion models in multi-channel transmitters has encountered many problems. For example, the complexity of the traditional model increases due to the increase in the number of channels, and the performance of the traditional model deteriorates due to the crosstalk between different channels and the inconsistency of the PA behavior of different channels. Summary of the invention

[0003] The present disclosure provides a multi-channel transmitter, a communication device and a distortion compensation method.

[0004] In a first aspect, an embodiment of the present disclosure provides a multi-channel transmitter, including a first pre-distortion module, a second pre-distortion module and a processing module, wherein the first pre-distortion module includes a first pre-distortion model, the second pre-distortion module includes a second pre-distortion model, and the second pre-distortion model is generated based on the training of the first pre-distortion model;

[0005] The first predistortion model is used to perform first predistortion processing on input signals of multiple channels respectively, to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels; the other channels are other channels among the multiple channels except the designated channel;

[0006] The second predistortion model is used to perform a second predistortion process on at least the input signals of the multiple channels to obtain a second predistortion signal of each of the other channels;

[0007] The processing module is used for processing the second predistortion signal of each of the other channels and the first predistortion signal of each of the other channels to obtain the third predistortion signal of each of the other channels.

[0008] In a second aspect, an embodiment of the present disclosure further provides a communication device, comprising the multi-channel transmitter as described above.

[0009] In a third aspect, an embodiment of the present disclosure further provides a distortion compensation method, the method comprising:

[0010] Using a first predistortion model, respectively perform first predistortion processing on the input signals of the multiple channels to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels; the other channels are channels other than the designated channel;

[0011] Performing a second predistortion process on at least the input signals of the multiple channels using a second predistortion model to obtain a second predistortion signal of each of the other channels;

[0012] The second predistortion signal of each of the other channels and the first predistortion signal of each of the other channels are processed to obtain a third predistortion signal of each of the other channels.

[0013] In a fourth aspect, an embodiment of the present disclosure further provides a computer storage medium on which a computer program is stored, wherein the program, when executed, implements the distortion compensation method as described above.

[0014] The multi-channel transmitter provided by the embodiment of the present disclosure includes a first predistortion module, a second predistortion module and a processing module. The first predistortion module includes a first predistortion model, the second predistortion module includes a second predistortion model, and the second predistortion model is generated according to the training of the first predistortion model; the first predistortion model is used to perform a first predistortion process on input signals of multiple channels respectively to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels; the other channels are other channels among the multiple channels except the designated channel; the second predistortion model is used to perform a second predistortion process on at least the input signals of the multiple channels to obtain second predistortion signals of each other channel. True signal; the processing module is used to process the second predistortion signal of each other channel and the first predistortion signal of each other channel to obtain the third predistortion signal of each other channel; the embodiment of the present disclosure adopts a first predistortion model and a second predistortion model that are interconnected, and uses the first predistortion model to perform predistortion processing on the designated channel, wherein the first predistortion model is generated based on the distortion behavior modeling of the designated channel, and the second predistortion model is used to perform predistortion processing on the difference between other channels and the designated channel, and the second predistortion model is generated based on the difference modeling between other channels and the designated channel, thereby compensating for the inconsistency of the power amplifiers of multiple channels. The embodiment of the present disclosure can perform predistortion compensation on multiple channels at the same time, simplify the predistorter structure of the multi-channel transmitter, reduce the consumption of computing resources in the multi-channel scenario, and improve the accuracy of predistortion compensation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Module schematic diagram of a multi-channel transmitter provided in an embodiment of the present disclosure Figure 1 ;

[0016] Figure 2 Module schematic diagram of a multi-channel transmitter provided in an embodiment of the present disclosure Figure 2 ;

[0017] Figure 3 A schematic diagram of a multi-channel transmitter provided for an embodiment of the present disclosure;

[0018] Figure 4 A schematic diagram of the process of the distortion compensation method provided in the embodiment of the present disclosure Figure 1 ;

[0019] Figure 5 A schematic diagram of a process for training a first pre-distortion model provided in an embodiment of the present disclosure;

[0020] Figure 6 A schematic diagram of a process for training a second pre-distortion model provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] Example embodiments will be described more fully below with reference to the accompanying drawings, but the example embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. On the contrary, the purpose of providing these embodiments is to make the present disclosure thorough and complete and to enable those skilled in the art to fully understand the scope of the present disclosure.

[0022] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0023] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded.

[0024] The embodiments described herein may be described with reference to plan views and / or cross-sectional views by means of idealized schematic diagrams of the present disclosure. Therefore, the example illustrations may be modified according to manufacturing techniques and / or tolerances. Therefore, the embodiments are not limited to the embodiments shown in the accompanying drawings, but include modifications of the configurations formed based on the manufacturing process. Therefore, the regions illustrated in the accompanying drawings have schematic properties, and the shapes of the regions shown in the figures illustrate the specific shapes of the regions of the elements, but are not intended to be limiting.

[0025] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless explicitly defined as such herein.

[0026] Traditional digital pre-distortion models are derived based on Volterra series, such as MP (Memory Polynomial) model, GMP (Generalized Memory Polynomial) model, DVR (Decomposition Vector Rotation) model, etc., which are mostly applicable to single-channel transmitters. In 5G communication systems, multi-channel transmitter technology has gradually matured, and there is an urgent need for a new digital pre-distortion model that can jointly compensate for the nonlinear distortion of multiple channels. In actual situations, due to the existence of non-ideal factors, the nonlinear state of PAs between different channels is inconsistent, and it is difficult to find explicit mathematical laws to describe such inconsistencies. If the traditional digital pre-distortion model is still used, it is difficult to use one model to simultaneously compensate for the nonlinearity of a large number of MIMO arrays. Only one digital pre-distortion model can be deployed for each channel. However, for transmitters with large-scale MIMO arrays with a large number of channels, the hardware complexity and resource consumption are huge.

[0027] Neural networks have shown excellent nonlinear fitting capabilities and can be used as a technical means to simultaneously fit the PA inconsistency between different channels. At present, there are two main types of nonlinear distortion compensation schemes for all-digital MIMO systems. The first type is based on improvements to traditional models. For example, zero-forcing precoding and predistorter cascade methods are used to reduce the model size, or, in order to reduce the order of traditional models, a closed-loop decorrelation algorithm and a model estimation method using small bandwidth observations are proposed. The second type is a multi-channel predistortion algorithm based on neural networks. For example, neural networks are used to simultaneously train the nonlinearity of multiple channels, trying to solve the I / Q signal imbalance and crosstalk problems between multiple channels, which can improve the nonlinear distortion caused by multi-channel crosstalk at a lower complexity; or, neural networks are used to uniformly train signals in different beam directions during beam scanning of multi-channel chips. However, the above nonlinear distortion compensation schemes all compensate for the crosstalk between multiple channels, but do not compensate for the inconsistency of the PA states of each channel, and the latter is the more serious distortion problem in the actual all-digital architecture multi-channel transmitter.

[0028] In order to solve the above problems, the present disclosure provides a multi-channel transmitter, such as Figure 1 As shown, the multi-channel transmitter includes a first predistortion module 1, a second predistortion module 2 and a processing module 3. The first predistortion module 1 includes a first predistortion model, the second predistortion module 2 includes a second predistortion model, and the second predistortion model is generated by training the first predistortion model. The first predistortion model is used to perform a first predistortion process on the input signals of multiple channels respectively to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels; other channels are other channels among the multiple channels except the designated channel. The second predistortion model is used to perform a second predistortion process on at least the input signals of multiple channels to obtain a second predistortion signal of each other channel. The processing module 3 is used to process the second predistortion signal of each other channel and the first predistortion signal of each other channel to obtain a third predistortion signal of each other channel.

[0029] The designated channel is any one of the channels in the multi-channel transmitter, and the channels other than the designated channel are called other channels.

[0030] The first pre-distortion model can perform nonlinear distortion compensation on the designated channel, and the second pre-distortion model is obtained by modeling the difference between each other channel and the designated channel.

[0031] The disclosed embodiment adopts a first pre-distortion model and a second pre-distortion model that are interconnected, uses the first pre-distortion model to perform distortion compensation on a designated channel, and uses the second pre-distortion model to perform distortion compensation on the difference between other channels and the designated channel, wherein the first pre-distortion model is generated based on the distortion behavior modeling of the designated channel, and the second pre-distortion model is generated based on the difference modeling between other channels and the designated channel, thereby compensating for PA inconsistency of multiple channels. The disclosed embodiment can perform pre-distortion compensation on multiple channels simultaneously, simplify the pre-distorter structure of a multi-channel transmitter, reduce the consumption of computing resources in a multi-channel scenario, and improve the accuracy of pre-distortion compensation.

[0032] The inventors have found that the nonlinear distortion of the multi-channel transmitter is mainly caused by the power amplifiers of each channel. Figure 2 and Figure 3 As shown, the multi-channel transmitter further includes a power amplifier module 4, which is connected to the first predistortion module 1 and the processing module 2. The power amplifier module 4 includes a plurality of power amplifiers, each of which corresponds to one channel.

[0033] In some embodiments, Figure 3As shown, taking three channels as an example, the power amplifier module 4 includes three power amplifiers PA1-PA3, wherein the first channel is a designated channel, the power amplifier of the designated channel is PA1, the second channel and the third channel are other channels, the power amplifier of the second channel is PA2, and the power amplifier of the third channel is PA3.

[0034] Correspondingly, each of the power amplifiers corresponding to the other channels is used to generate the first output signal of the channel according to the third predistortion signal of the corresponding channel; the power amplifier corresponding to the designated channel is used to generate the first output signal of the designated channel according to the first predistortion signal of the designated channel. In other words, the power amplifiers of each channel respectively amplify the predistortion signal input to the power amplifier of the corresponding channel to obtain the first output signal of the corresponding channel, and the first output signal is the PA output signal. Among them, the first output signal of the designated channel is generated by amplifying the first predistortion signal, and the first output signal of each other channel is generated by amplifying the third predistortion signal of each other channel.

[0035] In some embodiments, the second predistortion model is used to obtain the third predistortion signal of each other channel, and perform the second predistortion processing on the third predistortion signal of each other channel and the input signal of multiple channels to obtain the second predistortion signal of each other channel. That is, the third predistortion signal of each other channel output by the processing module 3 can be input into the second predistortion module 2 as one of the input signals of the second predistortion module 2.

[0036] The second predistortion model is used to obtain the amplitude and / or phase of the input signals of multiple channels, and perform a second predistortion process on the input signals of multiple channels and the amplitude and / or phase of the input signals of multiple channels to obtain the second predistortion signal of each other channel, or perform a second predistortion process on the input signals of multiple channels, the amplitude and / or phase of the input signals of multiple channels and the third predistortion signal of multiple other channels to obtain the second predistortion signal of each other channel. It should be noted that the input signals of multiple channels are digital signals having real and imaginary parts, and the amplitude and phase are generated based on the input signals, that is, the input signals are operated to obtain the amplitude and phase, and the amplitude and phase are both digital signals. That is to say, the second pre-distortion model can use two types of signals, namely, input signals of multiple channels and amplitude (or phase) of the input signals, as inputs to the second pre-distortion model; it can also use three types of signals, namely, input signals of multiple channels, amplitude and phase of the input signals, as inputs to the second pre-distortion model; it can also use four types of signals, namely, input signals of multiple channels, amplitude and phase of the input signals, and third pre-distortion signals of other channels, as inputs to the second pre-distortion model, so as to improve the pre-distortion compensation accuracy of the second pre-distortion model.

[0037] Due to the memory effect of the power amplifier, the output signal at the current moment is not only related to the input signal at the current moment, but also to the input signal in the previous period of time. In order to compensate for the memory effect of the power amplifier, it is necessary to use the historical signal as the input signal of the pre-distortion model. Therefore, in some embodiments, the input signals of multiple channels are the input signals of multiple channels at the current moment and within the previous preset time length. The amplitude and / or phase of the input signals of multiple channels are the amplitude and / or phase of the input signals of multiple channels at the current moment and within the previous preset time length. The third pre-distortion signal of each other channel is the third pre-distortion signal of each other channel within the preset time length before the current moment.

[0038] That is, in some embodiments, the second predistortion model is used to obtain the third predistortion signal of each other channel within a previously preset time length, and perform a second predistortion process on the third predistortion signal of each other channel within the previously preset time length and the input signals of multiple channels at the current moment and within the previously preset time length, to obtain the second predistortion signal of each other channel at the current moment.

[0039] In some embodiments, the second predistortion model is used to obtain the amplitude and / or phase of the input signals of multiple channels, and perform a second predistortion processing on the input signals of multiple channels at the current moment and within a previous preset time length and the amplitude and / or phase of the input signals to obtain the second predistortion signal of each other channel at the current moment; or, perform a second predistortion processing on the input signals of multiple channels at the current moment and within a previous preset time length and the amplitude and / or phase of the input signals, and the third predistortion signal of each other channel within the previous preset time length to obtain the second predistortion signal of each other channel at the current moment.

[0040] In some embodiments, the processing module 3 is used to, for each other channel, superimpose the second predistortion signal of the other channel with the first predistortion signal of the other channel to obtain a third predistortion signal of the other channel. Figure 3 As shown, the processing module 3 may be an adder. The first predistortion signal of each other channel output by the first predistortion model is added to the second predistortion signal of each other channel output by the second predistortion model to obtain a third predistortion signal of each other channel, and the third predistortion signal of each other channel is respectively transmitted to the corresponding power amplifier of each other channel to realize nonlinear distortion compensation of a multi-channel transmitter.

[0041] In some embodiments, the first pre-distortion model may be a model based on Volterra series or a neural network model. The second pre-distortion model may be a model based on Volterra series or a neural network model. Among them, the model based on Volterra series may include but is not limited to: generalized memory polynomial model (GMP), memory polynomial model (MP), decomposition vector rotation model (DVR), piecewise linearization model (CPWL). The neural network model may include but is not limited to: a multi-layer fully connected neural network model, a convolutional neural network model, and a recurrent neural network model.

[0042] It should be noted that the types of the first pre-distortion model and the second pre-distortion model can be arbitrarily combined, wherein at least one pre-distortion model adopts a neural network model, which has a better pre-distortion compensation effect. In the embodiment of the present disclosure, the first pre-distortion model is a model based on Volterra series, and the second pre-distortion model is a neural network model.

[0043] The disclosed embodiment is different from the current multi-channel DPD technology. It adopts a scheme of connecting the Volterra model with the neural network model, uses the neural network and the traditional polynomial model in combination, and makes full use of the advantages of different types of models to improve the performance of the digital predistorter. Among them, the first predistortion model uses the GMP model, and the GMP model is responsible for modeling the selected single channel (i.e., the specified channel). After the GMP model is modeled using the specified channel, the neural network model is modeled using the differences between other channels and the specified channel.

[0044] The input signal of the second predistortion model may be input in a serial manner or in a parallel manner. The transmission manners of the two input signals are described below respectively.

[0045] In some embodiments, the input signals of each other channel are input into the neural network model of the second predistortion module in a serial manner. The multi-channel transmitter has a total of 16 transmission channels. The input signals of multiple channels first pass through the parallel-serial conversion module, and the parallel-serial conversion module connects the input signals of each channel in chronological order, and converts the 16-way parallel input signals with a low rate into a single-channel serial input signal with a high rate. After the single-channel serial input signal passes through the second predistortion module, a single-channel serial predistortion signal is obtained, and then the high-rate serial signal is restored to a low-rate 16-way parallel predistortion signal through the serial-parallel conversion module. The 16-way parallel predistortion signals are respectively input into the PA of the corresponding channels to obtain the PA output signals of multiple channels. It should be noted that during the training process of the first predistortion model and the second predistortion model, the PA output signals of multiple channels also pass through the parallel-serial conversion module in a similar manner, and are converted from the low-rate 16-way parallel PA output signals to the high-rate single-channel serial PA output signals.

[0046] The second predistortion model has 46 input signals (digital signals), which are: (1) the real part (I), imaginary part (Q), amplitude and phase of the single-channel serial input signal at the current moment and 1-7 moments before (i.e., within the preset time length), a total of 8*4=32 channels; (2) the real part (I) and imaginary part (Q) of the single-channel serial third predistortion signal at 1-7 moments before, a total of 7*2=14 channels. The second predistortion model has 2 output signals, which are: 15 channels of real and imaginary parts of the second predistortion signal.

[0047] In some embodiments, the input signals of other channels are input into the neural network model of the second predistortion module in parallel. The channel transmitter has a total of 16 transmission channels. Assume that the input signals of the second predistortion model have a total of 384 channels, namely: the real part (I), imaginary part (Q) and amplitude of the input signals of 16 channels at the current moment and the previous 1-7 moments (i.e., within the preset time length), a total of 16*8*3=384 channels. The output signals of the second predistortion model have a total of 30 channels, namely: the real part and imaginary part of the predistortion signals of 15 channels.

[0048] The embodiment of the present disclosure also provides a communication device, including the multi-channel transmitter as described above, and the communication device may be a multi-channel transmitter base station with a fully digital architecture.

[0049] The communication device provided by the embodiment of the present disclosure includes a multi-channel transmitter, which includes a processing module, a first predistortion module including a first predistortion model, and a second predistortion module including a second predistortion model, wherein the second predistortion model is generated according to the training of the first predistortion model; the first predistortion model is used to perform a first predistortion process on input signals of multiple channels respectively to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels; the other channels are other channels among the multiple channels except the designated channel; the second predistortion model is used to perform a second predistortion process on at least the input signals of the multiple channels to obtain second predistortion signals of each other channel. True signal; the processing module is used to process the second predistortion signal of each other channel and the first predistortion signal of each other channel to obtain the third predistortion signal of each other channel; the embodiment of the present disclosure adopts a first predistortion model and a second predistortion model that are interconnected, and uses the first predistortion model to perform distortion compensation on the designated channel, and uses the second predistortion model to perform distortion compensation on the difference between other channels and the designated channel, wherein the first predistortion model is generated based on the distortion behavior modeling of the designated channel, and the second predistortion model is generated based on the difference modeling between other channels and the designated channel, thereby compensating for the inconsistency of the power amplifiers of multiple channels. The embodiment of the present disclosure can perform predistortion compensation on multiple channels at the same time, simplify the structure of the predistorter of the multi-channel transmitter, reduce the consumption of computing resources in the multi-channel scenario, and improve the accuracy of predistortion compensation.

[0050] The embodiment of the present disclosure also provides a distortion compensation method, which is applied to the multi-channel transmitter as described above. Figure 4 and Figure 1-2 As shown, the method comprises the following steps:

[0051] Step S11 , performing first predistortion processing on input signals of multiple channels respectively using a first predistortion model to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels.

[0052] The other channels are channels other than the designated channel in the multi-channel transmitter. In this step, the first predistortion model in the first predistortion module performs the first predistortion processing on the input signals of the multiple channels to obtain the predistortion signals of the multiple channels. The designated channel uses the first predistortion model to complete the predistortion compensation, while the other channels need to perform predistortion compensation again according to the difference between them and the designated channel.

[0053] Step S12: Perform a second predistortion process on at least the input signals of the multiple channels using a second predistortion model to obtain second predistortion signals of the other channels.

[0054] In this step, the second predistortion model in the second predistortion module performs second predistortion processing on the input signals of multiple channels to obtain second predistortion signals of other channels.

[0055] Step S13: Process the second predistortion signal of each other channel and the first predistortion signal of each other channel to obtain a third predistortion signal of each other channel.

[0056] In this step, the processing module generates a third predistortion signal of each other channel according to the second predistortion signal and the first predistortion signal of the channel, thereby completing predistortion compensation of input signals of each other channel.

[0057] In some embodiments, for each other channel, the second predistortion signal of the other channel is superimposed with the first predistortion signal of the other channel to obtain a third predistortion signal of the other channel.

[0058] The distortion compensation method provided by the embodiment of the present disclosure performs a first predistortion processing on the input signals of multiple channels respectively to obtain the first predistortion signal of the designated channel and the first predistortion signal of other channels; performs a second predistortion processing on the input signals of at least multiple channels to obtain the second predistortion signal of each other channel; processes the second predistortion signal of each other channel and the first predistortion signal of each other channel to obtain the third predistortion signal of each other channel; performs distortion compensation on the designated channel using the first predistortion model, and performs distortion compensation on the difference between the other channels and the designated channel using the second predistortion model, and the second predistortion model for the second predistortion processing is generated based on the difference model between the other channels and the designated channel, thereby compensating for the inconsistency of the power amplifiers of multiple channels. The embodiment of the present disclosure can perform predistortion compensation on multiple channels at the same time, simplify the preprocessor structure of the multi-channel transmitter, reduce the consumption of computing resources in the multi-channel scenario, and improve the accuracy of predistortion compensation.

[0059] In some embodiments, the first predistortion model is trained according to an input signal of a designated channel and an output signal of the designated channel.

[0060] In some embodiments, the second predistortion model is trained based on input signals of multiple channels and error signals of each other channel, wherein the error signal of each other channel is calculated based on the first predistortion model, the input signal of each other channel and the output signal of each other channel, that is, the error signal can reflect the difference between the other channels and the specified channel.

[0061] In some embodiments, Figure 5 As shown, the steps of training the first pre-distortion model include:

[0062] Step S21, arbitrarily select one of the channels as a designated channel, input the input signal of the designated channel to the multi-channel transmitter, obtain the second output signal of the designated channel, and calculate the target predistortion signal of the designated channel according to the input signal of the designated channel and the second output signal of the designated channel.

[0063] The second output signal of the designated channel is the output signal of the power amplifier of the designated channel. In this step, an ILC (iterative learning control) method, a DL (Direct Learning) method or an IL (Indirect Learning) method may be used to calculate the target predistortion signal of the designated channel.

[0064] Step S22: training a pre-established initial first predistortion model to obtain first coefficients according to the input signal of the designated channel and the target predistortion signal of the designated channel, and generating a first predistortion model according to the first coefficients.

[0065] The pre-established initial first pre-distortion model may be a model based on Volterra series or a neural network model, wherein the model based on Volterra series may include but is not limited to: a generalized memory polynomial model (GMP), a memory polynomial model (MP), a decomposed vector rotation model (DVR), and a piecewise linearization model (CPWL). The neural network model may include but is not limited to: a multi-layer fully connected neural network model, a convolutional neural network model, and a recurrent neural network model.

[0066] After steps S21-S22, the training of the first pre-distortion model is completed. In the subsequent process of training the second pre-distortion model, it is still necessary to rely on the trained first pre-distortion model.

[0067] In some embodiments, Figure 6 As shown, the steps of training the second pre-distortion model include:

[0068] Step S31 : generating fourth predistortion signals of other channels according to the input signals of other channels and the first predistortion model.

[0069] In this step, the input signals of the other channels are input into the trained first predistortion model, and the trained first predistortion model outputs the fourth predistortion signals of the other channels.

[0070] Step S32: input the input signal of each other channel to the multi-channel transmitter to obtain the second output signal of each other channel, and calculate the target predistortion signal of each other channel according to the input signal of each other channel and the second output signal of each other channel.

[0071] In this step, the target predistortion signal of each other channel may be calculated by using the ILC method, the DL method or the IL method.

[0072] Step S33: for each other channel, calculate the error signal of the other channel according to the fourth predistortion signal of the other channel and the target predistortion signal of the other channel.

[0073] The target predistortion signals of the other channels are acquired in the same manner as the target predistortion signal of the designated channel, and can be acquired synchronously with the target predistortion signal of the designated channel.

[0074] In this step, for each other channel, a difference between the fourth predistortion signal of the other channel and the target predistortion signal of the other channel is calculated to obtain an error signal of the other channel.

[0075] Step S34: training a pre-established initial second predistortion model to obtain second coefficients at least according to the input signals of the multiple channels and the error signals of the other channels, and generating the second predistortion model according to the second coefficients.

[0076] In this step, at least the input signals of multiple channels are used as the input of the initial second pre-distortion model, and the error signals of other channels obtained in step S33 are used as the output of the initial second pre-distortion model. The initial second pre-distortion model can be trained by the DL method or the IL method. After the initial second pre-distortion model converges, the second coefficient can be obtained, and the second pre-distortion model is generated according to the second coefficient.

[0077] After steps S31 - S34 , the training of the second pre-distortion model is completed.

[0078] In some embodiments, the step of training a pre-established initial second predistortion model to obtain second coefficients (i.e., step S34) at least according to the input signals of the multiple channels and the error signals of each other channel comprises the following steps:

[0079] Step S341: input the input signals of multiple channels into a pre-established initial second predistortion model to obtain fifth predistortion signals of other channels.

[0080] It should be noted that the execution order of this step and steps S31-S33 is not limited.

[0081] Step S342: input the fourth predistortion signal of each other channel and the fifth predistortion signal of each other channel into a processing module to obtain a sixth predistortion signal of each other channel.

[0082] Step S343: training a pre-established initial second predistortion model to obtain second coefficients at least according to the input signals of the multiple channels, the sixth predistortion signals of the other channels, and the error signals of the other channels.

[0083] In this step, the input signals of multiple channels at the current moment and within the previous preset time length and the sixth pre-distortion signal of each other channel within the previous preset time length are used as the input of the initial second pre-distortion model, and the error signals of each other channel obtained in step S33 are used as the output of the initial second pre-distortion model. The initial second pre-distortion model can be trained by the DL method or the IL method. After the initial second pre-distortion model converges, the second coefficient can be obtained, and the second pre-distortion model is generated according to the second coefficient.

[0084] In some embodiments, the step of training a pre-established initial second pre-distortion model to obtain a second coefficient (i.e., step S343) at least according to the input signals of multiple channels, the sixth pre-distortion signals of each other channel, and the error signals of each other channel includes the following steps: training a pre-established initial second pre-distortion model to obtain a second coefficient according to the input signals of multiple channels, the error signals of each other channel, and the amplitude and / or phase of the input signals of multiple channels; or training a pre-established initial second pre-distortion model to obtain a second coefficient according to the input signals of multiple channels, the sixth pre-distortion signals of each other channel, the error signals of each other channel, and the amplitude and / or phase of the input signals of multiple channels. That is, the input signals of multiple channels and the amplitude and / or phase of the input signals of multiple channels can be used as the input of the initial second pre-distortion model, and the error signals of each other channel obtained in step S33 can be used as the output of the initial second pre-distortion model, and the initial second pre-distortion model can be trained by the DL method or the IL method, and the second coefficient can be obtained after the initial second pre-distortion model converges, and the second pre-distortion model is generated according to the second coefficient. The input signals of multiple channels, the amplitudes and / or phases of the input signals of multiple channels, and the sixth predistortion signals of each other channel may also be used as inputs of the initial second predistortion model, and the error signals of each other channel obtained in step S33 may be used as outputs of the initial second predistortion model. The initial second predistortion model may be trained using the DL method or the IL method, and the second coefficients may be obtained after the initial second predistortion model converges, and the second predistortion model may be generated according to the second coefficients.

[0085] Due to the memory effect of the power amplifier, the output signal at the current moment is not only related to the input signal at the current moment, but also to the input signal in the previous period of time. In this way, in the process of training the first predistortion model and the second predistortion model, the input signals of the initial first predistortion model and the initial second predistortion model also include the input signals in the period of time before the current moment. Therefore, in some embodiments, the input signals of multiple channels are the input signals of multiple channels at the current moment and within the previous preset time length; the amplitude and / or phase of the input signals of multiple channels are the amplitude and / or phase of the input signals of multiple channels at the current moment and within the previous preset time length; the sixth predistortion signal of each other channel is the sixth predistortion signal of each other channel within the preset time length before the current moment.

[0086] The embodiment of the present disclosure further provides a computer storage medium on which a computer program is stored, wherein when the program is executed, the distortion compensation method as described above is implemented.

[0087] In order to clearly illustrate the technical solution of the present disclosure, the following takes the first pre-distortion model as a GMP model and the second pre-distortion model as a neural network model as an example to illustrate the process of training the first pre-distortion model and training the second pre-distortion model.

[0088] Step 1: Input the input signal of each channel to each channel of the multi-channel transmitter respectively, collect the PA output signal of each channel, and calculate the target predistortion signal of each channel by using the ILC method according to the input signal of each channel and the PA output signal of each channel.

[0089] Step 2, arbitrarily select one of the channels in the transmitter, use the GMP model to build a model based on the input signal of the designated channel and the target predistortion signal, obtain an initial first predistortion model, train the initial first predistortion model, and obtain a first coefficient.

[0090] Step 3: Calculate the difference between the fourth predistortion signal of each other channel and the corresponding target predistortion signal to obtain the error signal of each other channel.

[0091] Step 4, use the neural network model to build a model and obtain an initial second pre-distortion model. The input signal of each channel at the current moment and within the previous preset time length, the amplitude and phase of the input signal of each channel at the current moment and within the previous preset time length, and the sixth pre-distortion signal of each other channel within the previous preset time length are used as the input of the initial second pre-distortion model, and the error signal of each other channel calculated in step 3 is used as the output of the initial second pre-distortion model, and the initial second pre-distortion model is trained to obtain the second coefficient. Among them, the sixth pre-distortion signal within the previous preset time length is obtained by the processing module processing the fifth pre-distortion signal of each other channel within the previous preset time length and the fourth pre-distortion signal of each other channel within the previous preset time length. The fifth pre-distortion signal within the previous preset time length is the output of the second pre-distortion model within the previous preset time length.

[0092] The disclosed embodiment makes full use of the excellent nonlinear modeling ability and generalization ability of the neural network, and cooperates with the traditional GMP model to effectively model the PA inconsistency between different channels, so that only one set of models needs to be configured to achieve the joint modeling of multiple channels, reducing the complexity of the digital pre-distortion module. The main feature of this set of models is to use the traditional GMP model and the neural network model in cascade, first use the GMP model to model a specified channel, and the solved GMP model will not change when used in other channels, and use the neural network to model the difference between other channels and the specified channel. In this way, the traditional model can improve the modeling accuracy of a single channel, and the neural network model can be used to uniformly model different channels. The function of nonlinear compensation for multiple channels at the same time is finally realized, and the multi-channel transmitter can maintain a relatively impressive performance. Compared with the traditional "one-to-one" solution of configuring a traditional polynomial model for each channel, the present invention can realize the "one-to-many" pre-distortion module configuration of a multi-channel transmitter.

[0093] Example embodiments have been disclosed herein, and although specific terms are employed, they are used and should be interpreted only in a general illustrative sense and not for limiting purposes. In some instances, it will be apparent to those skilled in the art that, unless otherwise expressly stated, features, characteristics, and / or elements described in conjunction with a particular embodiment may be used alone or in combination with features, characteristics, and / or elements described in conjunction with other embodiments. Therefore, it will be understood by those skilled in the art that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A multi-channel transmitter, characterized in that: The system comprises a first predistortion module, a second predistortion module and a processing module, wherein the first predistortion module comprises a first predistortion model, the second predistortion module comprises a second predistortion model, and the second predistortion model is generated by training the first predistortion model; The first predistortion model is used to perform first predistortion processing on input signals of multiple channels respectively, to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels; the other channels are other channels among the multiple channels except the designated channel; The second predistortion model is used to perform a second predistortion process on at least the input signals of the multiple channels to obtain a second predistortion signal of each of the other channels; The processing module is used for processing the second predistortion signal of each of the other channels and the first predistortion signal of each of the other channels to obtain the third predistortion signal of each of the other channels.

2. The multi-channel transmitter according to claim 1, characterized in that It also includes a power amplifier module, which is connected to the first predistortion module and the processing module. The power amplifier module includes a plurality of power amplifiers, and each power amplifier corresponds to a channel.

3. The multi-channel transmitter according to claim 2, characterized in that The second predistortion model is used to obtain the third predistortion signal of each of the other channels, and perform second predistortion processing on the third predistortion signal of each of the other channels and the input signals of the multiple channels to obtain the second predistortion signal of each of the other channels.

4. The multi-channel transmitter according to claim 3, characterized in that The second predistortion model is used to obtain the amplitude and / or phase of the input signals of the multiple channels, and perform second predistortion processing on the input signals of the multiple channels, the amplitude and / or phase of the input signals of the multiple channels, and the third predistortion signal of each of the other channels to obtain the second predistortion signal of each of the other channels.

5. The multi-channel transmitter according to claim 1, characterized in that The second predistortion model is used to obtain the amplitude and / or phase of the input signals of the multiple channels, and perform second predistortion processing on the input signals of the multiple channels and the amplitude and / or phase of the input signals of the multiple channels to obtain the second predistortion signals of each of the other channels.

6. The multi-channel transmitter according to claim 4 or 5, characterized in that: The input signals of the multiple channels are the input signals of the multiple channels at the current moment and within a preset time period before; the amplitudes and / or phases of the input signals of the multiple channels are the amplitudes and / or phases of the input signals of the multiple channels at the current moment and within a preset time period before; the third predistortion signals of each of the other channels are the third predistortion signals of each of the other channels within a preset time period before the current moment.

7. The multi-channel transmitter according to claim 1, characterized in that The processing module is used for, for each of the other channels, superimposing the second predistortion signal of the other channel with the first predistortion signal of the other channel to obtain a third predistortion signal of the other channel.

8. The multi-channel transmitter according to any one of claims 1 to 5 and 7, characterized in that: The first pre-distortion model is a model based on Volterra series or a neural network model; and / or the second pre-distortion model is a model based on Volterra series or a neural network model.

9. A communication device, characterized in that: Comprising the multi-channel transmitter as claimed in any one of claims 1-8.

10. A distortion compensation method, comprising: Using a first predistortion model, respectively perform first predistortion processing on input signals of multiple channels to obtain a first predistortion signal of a designated channel and first predistortion signals of other channels; The other channels are other channels among the multiple channels except the designated channel; Performing a second predistortion process on at least the input signals of the multiple channels using a second predistortion model to obtain a second predistortion signal of each of the other channels; The second predistortion signal of each of the other channels and the first predistortion signal of each of the other channels are processed to obtain a third predistortion signal of each of the other channels.

11. The distortion compensation method according to claim 10, characterized in that: The first predistortion model is trained according to the input signal of the designated channel and the output signal of the designated channel.

12. The distortion compensation method according to claim 11, characterized in that: The second pre-distortion model is trained based on the input signals of the multiple channels and the error signals of each of the other channels, wherein the error signals of each of the other channels are calculated based on the first pre-distortion model, the input signals of each of the other channels and the output signals of each of the other channels.

13. The distortion compensation method according to claim 12, characterized in that: The method further includes the step of training the first pre-distortion model, wherein the step of training the first pre-distortion model includes: Arbitrarily selecting one of the channels as a designated channel, inputting an input signal of the designated channel into the multi-channel transmitter, obtaining a second output signal of the designated channel, and calculating a target predistortion signal of the designated channel according to the input signal of the designated channel and the second output signal of the designated channel; According to the input signal of the designated channel and the target predistortion signal of the designated channel, a pre-established initial first predistortion model is trained to obtain first coefficients, and the first predistortion model is generated according to the first coefficients.

14. The distortion compensation method according to claim 13, characterized in that: The method further includes the step of training the second pre-distortion model, wherein the step of training the second pre-distortion model includes: generating a fourth predistortion signal of each of the other channels according to the input signal of each of the other channels and the first predistortion model; Inputting the input signal of each of the other channels into the multi-channel transmitter to obtain a second output signal of each of the other channels, and calculating a target predistortion signal of each of the other channels according to the input signal of each of the other channels and the second output signal of each of the other channels; For each of the other channels, calculating an error signal of the other channel according to a fourth predistortion signal of the other channel and a target predistortion signal of the other channel; At least according to the input signals of the multiple channels and the error signals of each of the other channels, a pre-established initial second pre-distortion model is trained to obtain second coefficients, and the second pre-distortion model is generated according to the second coefficients.

15. The distortion compensation method according to claim 14, characterized in that: The step of training a pre-established initial second predistortion model to obtain a second coefficient at least according to the input signals of the plurality of channels and the error signals of each of the other channels comprises: Inputting the input signals of the multiple channels into a pre-established initial second predistortion model to obtain fifth predistortion signals of the other channels; Inputting the fourth predistortion signal of each of the other channels and the fifth predistortion signal of each of the other channels into the processing module to obtain a sixth predistortion signal of each of the other channels; A pre-established initial second predistortion model is trained to obtain second coefficients at least according to the input signals of the multiple channels, the sixth predistortion signals of the other channels and the error signals of the other channels.

16. The distortion compensation method according to claim 15, characterized in that: The step of training a pre-established initial second predistortion model to obtain a second coefficient based on at least the input signals of the multiple channels, the sixth predistortion signals of the other channels, and the error signals of the other channels comprises: A pre-established initial second predistortion model is trained to obtain second coefficients according to the input signals of the multiple channels, the sixth predistortion signals of the other channels, the error signals of the other channels, and the amplitudes and / or phases of the input signals of the multiple channels.

17. The distortion compensation method according to claim 14, characterized in that: The step of training a pre-established initial second predistortion model to obtain a second coefficient at least according to the input signals of the plurality of channels and the error signals of each of the other channels comprises: According to the input signals of the multiple channels, the error signals of the other channels, and the amplitudes and / or phases of the input signals of the multiple channels, a pre-established initial second predistortion model is trained to obtain second coefficients.

18. The distortion compensation method according to claim 16, characterized in that: The input signals of the multiple channels are the input signals of the multiple channels at the current moment and within a preset time period before; the amplitudes and / or phases of the input signals of the multiple channels are the amplitudes and / or phases of the input signals of the multiple channels at the current moment and within a preset time period before; the sixth predistortion signal of each of the other channels is the sixth predistortion signal of each of the other channels within the preset time period before the current moment.

19. The distortion compensation method according to any one of claims 10 to 18, characterized in that: The processing of the second predistortion signal of each of the other channels and the first predistortion signal of each of the other channels to obtain a third predistortion signal of each of the other channels includes: For each of the other channels, the second predistortion signal of the other channel is superimposed with the first predistortion signal of the other channel to obtain a third predistortion signal of the other channel.

Images