A 5G carrier communication signal processing method

Abstract

The application discloses a 5G carrier communication signal processing method, and relates to the technical field of wireless communication signal processing. The method utilizes an outer loop phase estimation value to perform phase pre-correction on a baseband transmitting signal, performs nonlinear pre-distortion processing on the pre-correction signal based on a generalized memory polynomial model, and transmits the pre-correction signal through a radio frequency front end. A reference signal is obtained by collecting the output of a power amplifier through a feedback coupling branch. An inner loop residual error is calculated, and a phase error observation is extracted, which is used as an observation input to perform state estimation on the local oscillator phase noise by using an extended Kalman filter. The estimation result is fed back to be used for subsequent phase pre-correction, forming a closed-loop processing architecture in which the inner loop nonlinear pre-distortion and the outer loop phase noise estimation are mutually fed back. The hardware operation amount can be reduced in combination with an asymmetric multi-rate processing architecture. The application can dynamically decouple and compensate the coupling distortion of the nonlinear distortion of the power amplifier and the local oscillator phase noise.

Description

Technical Field

[0001] This invention relates to the field of wireless communication signal processing technology, specifically to a 5G carrier communication signal processing method for jointly adaptively compensating for nonlinear distortion of radio frequency power amplifiers and local oscillator phase noise in broadband multi-carrier aggregation scenarios of fifth-generation mobile communication and its evolution systems. It belongs to the technical field of 5G new waveforms and 5G carrier signal processing and can be used in the baseband signal processing system of next-generation mobile communication base stations and terminal equipment. Background Technology

[0002] With the deepening commercial deployment of fifth-generation mobile communication technology and next-generation mobile communication technologies evolving towards 5G, multi-carrier aggregation technology and various new 5G waveforms adapted to low peak-to-average power ratio requirements are widely adopted to meet the service demands for high bandwidth and high throughput. These include discrete Fourier transform extended orthogonal frequency division multiplexing waveforms with single-carrier frequency domain equalization characteristics, cyclic prefix orthogonal frequency division multiplexing waveforms, and related waveforms using frequency domain spectrum shaping technology. These new waveforms have certain advantages in improving spectral efficiency and reducing the peak-to-average power ratio of the signal envelope. However, in multi-carrier aggregation scenarios, due to the multiple component carriers... When waves are superimposed in the time domain, the peak-to-average power ratio of the signal envelope remains at a high level. In the radio frequency transmission channels of base stations and terminal equipment, the high peak-to-average power ratio makes it easy for the baseband signal to enter the nonlinear operating region of the power amplifier during the power boosting process. This leads to in-band distortion and out-of-band spectral leakage in the transmitted signal, which in turn worsens the modulation quality index of error vector amplitude. At the same time, as the communication operating frequency band evolves to higher frequency bands, the phase noise level generated by the local oscillator circuit also increases significantly. The local oscillator phase noise will cause random disturbances to the phase of the transmitted signal, further affecting the modulation accuracy.

[0003] In existing baseband signal processing technologies, digital predistortion is typically used to compensate for the nonlinear distortion of power amplifiers. Its basic principle is to pre-apply a distortion component to the signal in the baseband digital domain that is opposite to the nonlinear characteristics of the power amplifier, thus counteracting the nonlinear effects generated by the power amplifier during actual operation. Phase noise compensation is usually treated as a separate processing module independent of digital predistortion, used to separately estimate and correct the phase jitter introduced by the local oscillator. In typical existing implementations, the transmitter baseband processing unit first uses a digital predistortion model based on memory polynomials or reduced-order Volterra series to perform nonlinear predistortion on the baseband signal. In the true processing, the coefficients of the model are iteratively updated using an adaptive algorithm based on the error between the power amplifier output feedback signal and the ideal signal. Subsequently, in the independently set phase noise compensation module, a phase tracking algorithm is used to separately estimate and correct the phase jitter introduced by the local oscillator. The above two processing modules work in cascade in sequence in the signal processing link, and there is no information interaction between them. Those skilled in the art generally believe that if feedback coupling is generated between the digital predistortion and phase noise compensation modules, the mutual interference of the two types of error signals will cause the adaptive algorithm to converge unstably or even cause self-excited oscillation. Therefore, the existing technology generally avoids the design of information interaction between the two.

[0004] The aforementioned existing technical solutions can achieve certain compensation effects in scenarios with narrow signal bandwidth and a small number of carriers. However, in the broadband multi-carrier aggregation scenarios of 5G and its evolution systems, the nonlinear distortion of the power amplifier and the phase noise of the local oscillator will mutually modulate on the same time-frequency resources, forming deep cross-coupled distortion components. Since the existing digital predistortion module and phase noise compensation module each independently model only a single distortion source and there is no information interaction between them, they cannot identify and compensate for the above-mentioned cross-coupled distortion components. As a result, there is always an uneliminable coupled distortion floor noise in the residual error of the system, causing the performance of error vector amplitude to deteriorate and the out-of-band spectrum leakage to exceed the standard. To compensate for this defect, the existing technology usually adopts the method of increasing the order of the digital predistortion polynomial model and expanding the memory depth to indirectly cover the coupled distortion components. However, this method will cause the dimension of the model parameters to be updated to increase sharply and the computational complexity to increase significantly. In high sampling rate scenarios that need to meet the requirements of real-time processing, it is difficult to implement in real time on the existing base station field-programmable gate array or dedicated integrated circuit hardware platform. It may even cause self-excited oscillation due to the unstable convergence of the adaptive algorithm. Summary of the Invention

[0005] The technical problem this invention aims to solve is that in the broadband multi-carrier aggregation scenario of 5G and its evolution systems, there is a deep cross-coupling effect between the nonlinear distortion of the RF power amplifier and the phase noise of the local oscillator. Existing digital predistortion schemes and phase noise compensation schemes are independent of each other and have no information interaction, making it impossible to effectively decouple and compensate for this coupled distortion. This leads to deterioration of error vector amplitude performance and out-of-band spectrum leakage. At the same time, if the coupled distortion is indirectly covered by increasing the order of nonlinear modeling, the computational complexity will increase sharply, making it difficult to implement in real time on existing base station hardware platforms. This invention aims to provide a 5G carrier communication signal processing method that can perform joint adaptive decoupling compensation for the coupled distortion of the power amplifier nonlinear distortion and the local oscillator phase noise, while also meeting the requirements of low-complexity hardware real-time processing.

[0006] To address the aforementioned technical problems, the 5G carrier communication signal processing method provided by this invention includes: acquiring a baseband transmission signal; performing phase pre-correction on the baseband transmission signal using an outer loop phase estimate to obtain a pre-corrected baseband signal; performing nonlinear pre-distortion processing on the pre-corrected baseband signal based on a generalized memory polynomial model to obtain a pre-distorted output signal; amplifying the pre-distorted output signal through a power amplifier in a radio frequency front-end and modulating it with a local oscillator before transmission; acquiring a reference signal by collecting the output signal of the power amplifier through a feedback coupling branch; calculating the inner loop residual error based on the pre-distorted output signal and the reference signal; performing phase angle calculation on the conjugate product of the inner loop residual error and the pre-distorted output signal to obtain a phase error observation; using the phase error observation as an observation input; performing state estimation on the local oscillator phase noise using an extended Kalman filter to obtain an updated outer loop phase estimate; and feeding back the updated outer loop phase estimate to perform phase pre-correction on the baseband transmission signal at subsequent times, thereby forming a closed-loop processing process in which the inner loop nonlinear pre-distortion processing and the outer loop phase noise estimation processing mutually feedback each other.

[0007] This invention breaks through the conventional understanding that digital predistortion and phase noise compensation must be independent of each other and avoid information interaction to prevent convergence instability. By using the phase angle of the conjugate product of the inner-loop residual error and the predistortion signal as the outer-loop observation, a specific information extraction method is employed. This achieves effective information interaction between the inner and outer loops while ensuring stable closed-loop convergence. The above technical solution constructs a dual-loop joint adaptive compensation architecture with mutual feedback between the inner-loop equivalent baseband nonlinear equalization processing and the outer-loop phase noise extraction processing. The phase error observation obtained by extracting the residual error signal generated by the inner-loop predistortion processing serves as the observation update basis for the outer-loop extended Kalman filter state estimation, realizing the interaction between the power amplifier's nonlinear distortion component and the predistortion signal. The joint modeling of the phase noise components allows the modeling error of the inner loop model and the estimation error of the outer loop filter to correct each other during the closed-loop feedback process, thereby dynamically decoupling and compensating for the coupling distortion components of the two. Based on this, the present invention further provides an asymmetric multi-rate processing scheme, which downsamples the phase noise estimation operation with a relatively low rate of change to the low-rate domain for processing, while retaining the nonlinear predistortion multiplication and accumulation operation with a high rate of change in the high-rate domain consistent with the baseband sampling rate for processing. The signal connection between the high-rate domain and the low-rate domain is realized through interpolation reconstruction. Without changing the accuracy of the inner loop nonlinear modeling and the outer loop phase noise tracking accuracy, the overall hardware computation is significantly reduced.

[0008] The beneficial technical effects achieved by this invention include:

[0009] 1. This invention, through a dual-loop joint adaptive compensation architecture with mutual feedback between the inner and outer loops, can dynamically decouple the coupling distortion term between the power amplifier nonlinearity and the local oscillator phase noise, eliminate the coupling error floor noise that existing independent compensation schemes cannot handle, effectively improve the error vector amplitude performance of the transmitted signal, and suppress out-of-band spectrum leakage;

[0010] 2. This invention uses an asymmetric multi-rate processing architecture to perform phase noise estimation related operations in the low-rate domain, while only retaining the multiply-accumulate operations of nonlinear predistortion in the high-rate domain. Without significantly affecting the compensation performance, it can significantly reduce the hardware computational complexity, making the processing method provided by this invention easy to deploy in real time on existing base station field-programmable gate arrays or dedicated integrated circuit hardware platforms.

[0011] 3. The processing method provided by this invention is implemented entirely in the baseband digital domain, without the need to add additional hardware detection circuits or feedback channels to the RF front end. It has good compatibility with existing baseband processing architectures and is easy to implement in engineering. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating the overall architecture of the 5G carrier communication signal processing system of the present invention.

[0013] Figure 2 This is a flowchart of the inner loop nonlinear predistortion processing of the present invention;

[0014] Figure 3 This is a flowchart of the outer loop phase noise estimation and dual-loop collaborative decoupling process of the present invention. Detailed Implementation

[0015] The following is in conjunction with the appendix Figure 1-3 The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0016] I. System Overall Architecture

[0017] The 5G carrier communication signal processing method provided in this embodiment is applied to a processing system including a baseband signal transmitting unit, a dual-ring joint adaptive compensation unit, a radio frequency front-end unit, and a feedback coupling branch. The radio frequency front-end unit includes a local oscillator and a power amplifier. The dual-ring joint adaptive compensation unit is further subdivided into an inner-ring module and an outer-ring module. The inner-ring module includes a phase pre-rotation unit, a nonlinear pre-distortion unit based on a generalized memory polynomial, a coefficient adaptive update unit, and a residual error calculation unit. The outer-ring module includes a low-rate downsampling unit, a phase estimation unit based on an extended Kalman filter, and an interpolation reconstruction unit. The module operates in the high-rate domain consistent with the baseband sampling rate, while the outer loop module operates in the low-rate domain after downsampling. The phase error observation output by the inner loop residual error calculation unit is downsampled and sent to the outer loop phase estimation unit. The phase estimate output by the outer loop phase estimation unit is interpolated and reconstructed before being fed back to the inner loop phase pre-rotation unit, thus forming a complete double-loop closed feedback loop. The signal of the inner loop after pre-distortion processing is amplified by the power amplifier in the RF front-end unit and modulated by the local oscillator before being transmitted. At the same time, the actual output signal of the power amplifier is acquired through the feedback coupling branch and used by the inner loop coefficient adaptive update unit to calculate the residual error.

[0018] II. Inner Loop: Equivalent Baseband Nonlinear Predistortion Compensation Based on Generalized Memory Polynomial

[0019] The inner loop module uses a generalized memory polynomial model to model the equivalent baseband of the nonlinear memory effect of the power amplifier. Let the pre-corrected baseband signal obtained after the outer loop phase pre-correction be... The predistorted output signal of the inner loop nonlinear predistortion processing output Satisfy the following mathematical relationship:

[0020] ;

[0021] In the formula, Indicates the discrete-time sampling point number. This represents the upper bound of the nonlinear order of the generalized memory polynomial model, used to determine the range of nonlinear powers considered by the model. Indicates the first The memory depth corresponding to the order nonlinear term is used to characterize the number of historical sampling points considered for that order nonlinear term. The coefficients of the generalized memory polynomial model to be adaptively updated are used to characterize the power amplifier in the _th_ ... Order nonlinearity, delay Amplitude nonlinearity characteristics under individual sampling point conditions. This indicates the delay after outer loop phase pre-correction. The pre-calibrated baseband signal at each sampling point Indicates the amplitude of the corresponding signal envelope The power factor is used to characterize the amplitude nonlinear response of the power amplifier as the signal envelope changes. By assigning corresponding model coefficients to signal components with different delay depths and different nonlinear orders and performing weighted summation, the generalized memory polynomial model can accurately model the equivalent baseband characteristics of the power amplifier with memory effect with relatively low parameter dimensionality.

[0022] III. Outer Loop: Phase Noise Estimation and Tracking Based on Extended Kalman Filter

[0023] The outer loop module models the local oscillator phase noise as a first-order random walk process, constructs a state-space model, and uses an extended Kalman filter to recursively estimate the phase noise state. The state transition equation for the phase noise is:

[0024] ;

[0025] In the formula, Indicates the first The actual values ​​of the local oscillator phase noise at each low-rate sampling point. Indicates phase noise at the 1st The instantaneous angular frequency shift at each sampling point is used to characterize the drift rate of phase noise. This indicates the sampling period in the low-rate domain. This represents the process noise component with a mean of zero, used to characterize the uncertainty in the state transition process.

[0026] The prediction and update processes of the extended Kalman filter are represented as follows:

[0027] ;

[0028] ;

[0029] In the formula, Indicates based on the first Time information for the first The predicted estimate made based on the time phase. Indicates the first The optimal estimate at time [time]. , , These represent the estimation error covariance at the corresponding time points, used to characterize the degree of uncertainty in the phase estimation. Represents the process noise covariance. Represents the observation noise covariance. Indicates the first The Kalman gain at each low-rate sampling point is used to determine the correction weights for the observation pair's state update. Indicates the first The phase error observations extracted from the inner loop residual error at each low-rate sampling point are used to dynamically track and estimate the time-varying local oscillator phase noise through the recursive iteration of the above prediction and update process.

[0030] IV. Dual-ring collaborative decoupling mechanism

[0031] To achieve information exchange between the inner and outer loops and dynamic decoupling of coupling distortion, this implementation design includes the following closed-loop feedback processing mechanism.

[0032] First, a phase pre-rotation process is performed, and the phase estimate output by the outer loop module is... After being interpolated and reconstructed to a high-rate domain consistent with the baseband sampling rate, it is used to transmit the original baseband signal output by the baseband signal transmitting unit. Pre-calibration processing is performed, specifically satisfying the following relationship:

[0033] ;

[0034] In the formula, This represents the original high-speed baseband transmitted signal. This represents the high-rate phase noise estimate obtained after interpolation and reconstruction. This represents the compensation factor for the inverse rotation of the phase noise estimation components. This represents the pre-corrected baseband signal that is fed into the nonlinear predistortion unit after the phase noise estimation component has been eliminated.

[0035] Secondly, the inner loop residual error is calculated, and the predistortion output signal output by the inner loop nonlinear predistortion unit is used. The reference signal obtained after acquiring the actual output signal of the power amplifier through the feedback coupling branch. Subtracting the two, we obtain the inner loop residual error:

[0036] ;

[0037] The inner loop residual error It is also used for adaptive updating of coefficients in generalized memory polynomial models and extraction of outer loop observations, enabling information sharing between inner loop processing and outer loop processing.

[0038] Next, phase error observations are extracted, and information related to the phase coupling component is extracted from the inner loop residual error as the observation input for the outer loop extended Kalman filter:

[0039] ;

[0040] In the formula, This indicates the phase angle operation. This represents the conjugate of the inner-loop nonlinear predistortion output signal. This represents the extracted phase error observation, which contains coupling information between the power amplifier's nonlinear components and the local oscillator's phase noise components. This observation, after downsampling, is fed into the outer-loop extended Kalman filter module, corresponding to the observation in the aforementioned state-space model. .

[0041] Through the closed-loop processing link formed by the above-mentioned pre-correction, residual calculation, observation extraction, filtering estimation and feedback correction, the inner loop nonlinear modeling error and the outer loop phase noise estimation error are mutually corrected in the iterative process, thereby realizing the dynamic decoupling compensation of the coupling distortion between the nonlinear distortion component of the power amplifier and the local oscillator phase noise component. This is the core processing mechanism that distinguishes this invention from the existing independent compensation scheme.

[0042] V. Adaptive Update of Inner Loop Model Coefficients

[0043] The coefficients of the inner-loop generalized memory polynomial model are iteratively updated using a least-mean-square adaptive algorithm based on the inner-loop residual error. The specific update relationship is as follows:

[0044] ;

[0045] In the formula, , They represent the first The second iteration and the first The coefficients of the generalized memory polynomial model at the next iteration. This represents a pre-defined adaptive step size factor used to control the convergence speed and stability during the coefficient update process. Indicates delay The conjugate of the pre-corrected baseband signal at each sampling point, and the meanings of the other symbols are consistent with the above content. Through the above iterative update method based on residual error, the coefficients of the generalized memory polynomial model can be adaptively adjusted according to the changes in the nonlinear characteristics of the power amplifier and the coupling distortion state.

[0046] VI. Asymmetric Multi-rate Processing Architecture

[0047] Considering that the rate of change of local oscillator phase noise is relatively low compared to the baseband sampling rate required for nonlinear predistortion processing, this embodiment further provides an asymmetric multi-rate processing architecture. The outer loop phase noise estimation related operations are arranged in the low-rate domain, while the multiplication and accumulation operations of the inner loop nonlinear predistortion are retained in the high-rate domain. The high-rate processing domain operates at the original baseband sampling rate and only carries the multiplication and accumulation operations of nonlinear predistortion and the complex multiplication operations of phase pre-rotation. This part of the operation structure is relatively regular and is easy to implement using a field-programmable gate array or a pipelined operation array in an application-specific integrated circuit. The low-rate processing domain operates at the rate after downsampling and carries the state prediction and Kalman gain calculation operations in the outer loop extended Kalman filter. Since the operation frequency of this part is significantly reduced compared to the high-rate domain, the logic resources and power consumption it occupies are also reduced accordingly.

[0048] The downsampling and interpolation reconstruction processes between the high-rate domain and the low-rate domain satisfy the following mathematical relationships:

[0049] ;

[0050] ;

[0051] In the formula, This represents the downsampling factor, whose value is determined based on the ratio between the rate of change of the local oscillator phase noise and the baseband sampling rate. This represents the low-rate phase error observation sequence that has been downsampled and then fed into the outer-loop extended Kalman filter module. This represents the phase estimation sequence output by the outer-loop extended Kalman filter in the low-rate domain. This represents the phase estimate restored to the high-rate domain after linear interpolation reconstruction, used in the aforementioned phase pre-rotation process. Through the aforementioned asymmetric multi-rate processing architecture, the amount of hardware computation required for outer loop phase noise estimation can be significantly reduced without significantly affecting the compensation performance.

[0052] VII. Specific Implementation Process of the Processing Method

[0053] In the specific implementation, the initialization process is first performed, which sets the nonlinear order of the generalized memory polynomial model, the memory depth of each order, and the initial values ​​of the model coefficients. At the same time, the initial values ​​of the outer loop extended Kalman filter state, the initial value of the error covariance, the downsampling factor, the adaptive step size factor, the process noise covariance, the observation noise covariance, and the preset convergence threshold are set.

[0054] Subsequently, the current high-speed baseband transmission signal is read, and the phase estimate obtained by interpolation reconstruction at the previous moment is used to perform phase pre-correction processing on the baseband transmission signal to obtain the pre-corrected baseband signal. The pre-corrected baseband signal is substituted into the current generalized memory polynomial model to calculate the predistorted output signal, and the predistorted output signal is sent to the radio frequency transmission link for power amplification and local oscillator modulation transmission.

[0055] Next, the actual output signal of the power amplifier is acquired through the feedback coupling branch. After down-conversion and delay alignment processing, a reference signal is obtained. The predistorted output signal is subtracted from the reference signal to obtain the inner loop residual error. Based on the inner loop residual error, the coefficients of the generalized memory polynomial model are iteratively updated according to the aforementioned minimum mean square class adaptive update relationship, which is used for the predistortion processing of subsequent sampling points.

[0056] Furthermore, based on the conjugate of the inner loop residual error and the predistortion output signal, the phase error observation is extracted. After accumulating a certain number of high-rate sampling points, the phase error observation sequence is decimated once to obtain low-rate observations, which are then sent to the outer loop extended Kalman filter module. In the low-rate domain, the phase prediction value and prediction error covariance are first obtained based on the state estimate value of the previous moment. Then, the Kalman gain is calculated in combination with the current low-rate observation value to complete the state update and obtain the updated phase estimate value and update error covariance.

[0057] Finally, the updated low-rate phase estimation sequence is reconstructed into a high-rate phase estimation sequence consistent with the baseband sampling rate according to the aforementioned linear interpolation relationship, for use in the phase pre-correction processing of subsequent sampling points. At the same time, the energy of the inner loop residual error or its moving average is calculated. When the value is lower than the preset convergence threshold, the processing is determined to have entered a steady state and is switched to a steady-state tracking mode with a smaller adaptive step size for fine-tuning and updating. Otherwise, the subsequent sampling points are iteratively processed according to the above process, thus forming a continuously running closed-loop processing process.

[0058] VIII. Alternative Implementation Methods

[0059] In another embodiment of the present invention, the outer loop phase noise tracking part can use particle filtering instead of extended Kalman filtering to estimate the local oscillator phase noise state. This embodiment approximates the posterior probability distribution of the phase noise by using multiple randomly sampled particles. In scenarios where the local oscillator phase noise exhibits strong non-Gaussian and nonlinear characteristics, it can achieve tracking performance similar to or better than extended Kalman filtering. The bidirectional feedback closed-loop processing mechanism between it and the inner loop nonlinear predistortion unit is consistent with the aforementioned embodiment, that is, the phase error observation obtained by extracting the inner loop residual error is still used as the observation input, and the phase estimate value output by the outer loop is still fed back for the phase pre-correction processing of the inner loop.

[0060] In another embodiment of the present invention, the inner loop nonlinear modeling part can use a recurrent neural network model instead of a generalized memory polynomial model to perform data-driven modeling of the nonlinear memory effect of the power amplifier. This embodiment can also be combined with the aforementioned asymmetric multi-rate processing architecture, arranging the update operation of the neural network model weights in the low-rate domain, while retaining the multiplication and accumulation operations involved in the forward inference of the neural network in the high-rate domain, thereby controlling the complexity of hardware implementation while taking into account the accuracy of nonlinear modeling. In this embodiment, the phase pre-correction processing and the observation feedback processing mechanism between the inner and outer loops can both adopt the closed-loop processing design in the aforementioned embodiments.

[0061] The above-described specific embodiments are only some examples of the present invention. Those skilled in the art can make appropriate adjustments to the specific parameters of the inner loop nonlinear modeling method, the outer loop state estimation method, and the multi-rate processing architecture without departing from the concept of the present invention. All technical effects obtained by using the same or similar technical solutions as the present invention should fall within the protection scope of the present invention.

Claims

1. A 5G carrier communication signal processing method, characterized in that, The method includes: Acquire baseband transmission signal; The baseband transmitted signal is pre-corrected using the outer loop phase estimate to obtain the pre-corrected baseband signal; The pre-corrected baseband signal is subjected to nonlinear pre-distortion processing based on the generalized memory polynomial model to obtain the pre-distortion output signal. The predistorted output signal is amplified by the power amplifier of the radio frequency front end and modulated by the local oscillator before being transmitted. The output signal of the power amplifier is acquired through the feedback coupling branch to obtain the reference signal. Calculate the inner loop residual error based on the predistorted output signal and the reference signal; The phase error observation is obtained by performing a phase angle operation on the product of the inner loop residual error and the conjugate of the predistortion output signal. Using the phase error observation as the observation input, the local oscillator phase noise is estimated by extended Kalman filtering to obtain the updated outer loop phase estimate. The updated outer loop phase estimate is fed back to perform phase pre-correction on the baseband transmission signal at subsequent times, forming a closed-loop processing process in which the inner loop nonlinear pre-distortion processing and the outer loop phase noise estimation processing mutually feed back to each other.

2. The method according to claim 1, characterized in that, The nonlinear predistortion processing of the pre-calibrated baseband signal based on the generalized memory polynomial model includes: The pre-corrected baseband signal is multiplied by the product of the signal values ​​at different delay points and multiple powers of its signal envelope, and then multiplied by the corresponding model coefficients and summed to obtain the pre-distortion output signal. The model coefficients are used to characterize the amplitude nonlinearity of the power amplifier at different delay depths.

3. The method according to claim 1, characterized in that, The state estimation of local oscillator phase noise using extended Kalman filtering includes: The local oscillator phase noise is modeled as a first-order random walk process. Based on the phase estimate and angular frequency shift of the previous moment, the phase prediction and prediction error covariance of the current moment are predicted. Calculate the Kalman gain based on the phase error observation and the phase prediction value; The phase prediction value is corrected based on the Kalman gain to obtain the updated outer loop phase estimate and the update error covariance.

4. The method according to claim 1, characterized in that, The inner loop residual error is used to adaptively iteratively update the model coefficients of the generalized memory polynomial model, specifically as follows: In each iteration, the model coefficients are adjusted according to a preset step size factor based on the product of the current inner loop residual error, the conjugate of the pre-corrected baseband signal at the corresponding delay point, and the power of its signal envelope.

5. The method according to claim 1, characterized in that, Also includes: The phase error observations are downsampled to obtain a low-rate phase error observation sequence. The low-rate phase error observation sequence is then input into the extended Kalman filter for state estimation to obtain a low-rate outer loop phase estimation sequence. The low-rate outer loop phase estimation sequence is interpolated and reconstructed to obtain a high-rate phase estimation sequence consistent with the sampling rate of the baseband transmitted signal, and the high-rate phase estimation sequence is used to perform phase pre-correction on the baseband transmitted signal.

6. The method according to claim 5, characterized in that, The downsampling factor used to downsample the phase error observation is determined based on the ratio between the rate of change of the local oscillator phase noise and the sampling rate of the baseband transmitted signal.

7. The method according to claim 5, characterized in that, The interpolation reconstruction adopts a linear interpolation method, which calculates the phase estimate at the corresponding high-rate sampling point based on the phase estimates of two adjacent low-rate outer loops and the position ratio of the sampling point to be reconstructed relative to the adjacent low-rate sampling points.

8. The method according to claim 1, characterized in that, Also includes: The energy of the inner loop residual error is calculated. When the energy is lower than the preset convergence threshold, it is determined to enter the steady-state tracking mode. In the steady-state tracking mode, the model coefficients of the generalized memory polynomial model are updated with an adaptive step size smaller than the conventional iteration step size.

9. The method according to any one of claims 1 to 8, characterized in that, There is coupling distortion between the power amplifier and the local oscillator. The operation of extracting phase error observations is used to characterize the coupling information between the nonlinear distortion component of the power amplifier and the phase noise component of the local oscillator. The coupling distortion is dynamically decoupled and compensated through the closed-loop processing.

10. A 5G carrier communication signal processing device, characterized in that, It includes a baseband signal transmission unit, a dual-ring joint adaptive compensation unit, an RF front-end unit, and a feedback coupling branch; The radio frequency front-end unit includes a local oscillator and a power amplifier; The dual-loop joint adaptive compensation unit includes an inner loop module and an outer loop module. The inner loop module includes a phase pre-rotation unit, a nonlinear pre-distortion unit, a coefficient adaptive update unit, and a residual error calculation unit. The outer loop module includes a phase noise estimation unit. The phase pre-rotation unit is used to perform phase pre-correction on the baseband transmission signal output by the baseband signal transmission unit based on the phase estimate value output by the outer ring module, so as to obtain a pre-corrected baseband signal. The nonlinear predistortion unit is used to perform nonlinear predistortion processing on the pre-corrected baseband signal based on the generalized memory polynomial model to obtain a predistortion output signal. The predistortion output signal is amplified by the power amplifier of the radio frequency front-end unit and modulated by the local oscillator before being transmitted. The feedback coupling branch is used to acquire the output signal of the power amplifier to obtain a reference signal; The residual error calculation unit is used to calculate the inner loop residual error based on the predistortion output signal and the reference signal; The coefficient adaptive update unit is used to adaptively update the model coefficients of the nonlinear predistortion unit according to the inner loop residual error; The inner loop module is also used to extract phase error observations based on the conjugate of the inner loop residual error and the predistortion output signal, and output the phase error observations to the outer loop module; The phase noise estimation unit of the outer loop module is used to take the phase error observation as the observation input, use extended Kalman filtering to perform state estimation of the local oscillator phase noise, obtain the updated phase estimate, and feed the updated phase estimate back to the phase pre-rotation unit, forming a closed-loop processing architecture in which the inner loop nonlinear pre-distortion processing and the outer loop phase noise estimation processing mutually feed back to each other.

Images