Digital pre-distortion method and system for gallium nitride power amplifier

Abstract

The invention discloses a digital pre-distortion method and system for a gallium nitride power amplifier. The method comprises the following steps: calculating an instantaneous modulus value of an original signal, looking up a table to obtain a pre-distortion coefficient, carrying out pre-distortion compensation, inputting the pre-distortion compensation into the gallium nitride power amplifier, amplifying and outputting the pre-distortion compensation. The construction method of the predistortion coefficient lookup table comprises the following steps: acquiring an original signal and a feedback signal, and performing synchronous alignment processing; calculating an instantaneous module value corresponding to each sampling point in the synchronously aligned original signal, and averaging to obtain an average module value; constructing all items of the generalized memory polynomial model; adding a moving average generalized memory polynomial item in the generalized memory polynomial model to obtain a pre-distortion model; and estimating coefficients of all items in the pre-distortion model, and converting the coefficients into a pre-distortion coefficient lookup table which takes the instantaneous module value as an index. According to the invention, high-precision compensation for the signal distortion of the gallium nitride power amplifier is realized while high processing efficiency is achieved.

Description

Technical Field

[0001] This invention relates to the field of digital predistortion technology, and in particular to a digital predistortion method and system for gallium nitride power amplifiers. Background Technology

[0002] In radio frequency transceivers, small signals need to be amplified by a power amplifier (PA) before being transmitted. The signal amplified by the PA typically exhibits varying degrees of distortion. This distortion can be pre-compensated using digital pre-distortion (DPD) technology before the small signal enters the PA. After DPD compensation, the signal amplified by the PA results in a near-distortion-free linear amplification compared to the signal before DPD. Power amplifiers using gallium nitride (GaN) technology offer advantages such as higher gain, higher output power, and excellent thermal performance, better meeting the new requirements of modern electronics for high temperature, high power, high frequency, and radiation resistance. They are an ideal choice for 5G systems, and the long-term memory characteristics resulting from their trapping effect place new demands on DPDs.

[0003] Existing DPD schemes typically model the behavior of PA as GMP (Generalized Memory Polynomial), and the predistortion system structure based on this model is as follows: Figure 2 As shown, this model has a good compensation effect for LDMOS (Lateral Diffused Metal-Oxide-Semiconductor) type PAs, but it is often ineffective for GaN type PAs, especially in application scenarios with large signal power variations, such as (TDD Time Division Duplexing) scenarios. This model will cause the ACPR (Adjacent Channel Power Ratio) performance to fail to meet the requirements, or the EVM (Error Vector Magnitude) performance to deteriorate. Summary of the Invention

[0004] The main objective of this invention is to provide a digital predistortion method and system for gallium nitride power amplifiers, which achieves high-precision compensation for signal distortion of gallium nitride power amplifiers while having high processing efficiency.

[0005] The technical solution adopted in this invention is: a digital predistortion method for gallium nitride power amplifiers, comprising: The instantaneous magnitude of the original signal is calculated, the predistortion coefficient is obtained by looking up the table based on the instantaneous magnitude, predistortion compensation is performed, and then the signal is amplified and output by a gallium nitride power amplifier. The methods for constructing the predistortion coefficient lookup table include: The original signal before input to the gallium nitride power amplifier and the feedback signal output after amplification by the gallium nitride power amplifier are acquired and synchronized. The instantaneous magnitude value corresponding to each sampling point in the original signal after synchronization and alignment is calculated and averaged to obtain the average magnitude value. Based on the original signal after synchronization alignment and the instantaneous modulus values ​​corresponding to each sampling point, all terms of the generalized memory polynomial model are constructed. By introducing the average modulus value, a moving average generalized memory polynomial term is added to the generalized memory polynomial model to obtain the predistortion model. Based on the feedback signal after synchronization alignment, the coefficients of all terms in the predistortion model are estimated and converted into a predistortion coefficient lookup table indexed by the instantaneous modulus value.

[0006] According to the above technical solution, the feedback signal is acquired by extracting the feedback signal through coupling at the output of the gallium nitride power amplifier.

[0007] According to the above technical solution, the methods for synchronizing and aligning the original signal and the feedback signal include: Based on the acquired raw and feedback signals, a cross-correlation function is constructed, and the maximum cross-correlation function modulus and its corresponding time offset are found. The maximum cross-correlation function magnitude is compared with a preset threshold. If it is greater than the preset threshold, the synchronization is considered successful. The time offset corresponding to the maximum cross-correlation function magnitude is used as the time delay between the original signal and the feedback signal. The feedback signal is compensated for the time delay based on the time delay. If it is less than or equal to the preset threshold, the synchronization is considered unsuccessful. The original signal and the feedback signal are reacquired. The amplitude and phase of the feedback signal after time delay compensation are adjusted; wherein, the amplitude adjustment coefficient is related to the ratio of the magnitude of the maximum cross-correlation function to the power of the feedback signal, and the phase adjustment angle is the phase angle of the magnitude of the maximum cross-correlation function.

[0008] According to the above technical solution, the method for averaging the instantaneous magnitude values ​​corresponding to each sampling point in the original signal after synchronization alignment includes: For each sampling point, compare its instantaneous modulus with the average modulus corresponding to the previous sampling point. When the instantaneous modulus of the sampling point is greater than or equal to the average modulus of the previous sampling point, the average modulus of the sampling point is adjusted by the charging time constant; when the instantaneous modulus of the sampling point is less than the average modulus of the previous sampling point, the average modulus of the sampling point is adjusted by the discharging time constant.

[0009] According to the above technical solution, the output of the predistortion model includes the sum of a generalized memory polynomial term and a moving average generalized memory polynomial term; wherein, The generalized memory polynomial term is constructed based on the original signal after synchronization alignment and the instantaneous modulus value corresponding to each sampling point. Nonlinear transformation terms and cross-nonlinear transformation terms under different memory depths are constructed, and then each transformation term is multiplied by the corresponding model coefficients and summed. The moving average generalized memory polynomial term is constructed based on the original signal after synchronization and alignment and the instantaneous magnitude value corresponding to each sampling point. Nonlinear transformation terms and cross-nonlinear transformation terms at different memory depths are constructed. Each transformation term is multiplied by the corresponding model coefficients, and then multiplied by the average magnitude value. Finally, the terms are summed.

[0010] According to the above technical solution, the method for estimating the predistortion coefficient includes: Construct the regression matrix based on all terms of the predistortion model; The desired output signal is constructed based on the error between the original signal after synchronization and the feedback signal; The predistortion coefficients are obtained by solving the linear system formed by the regression matrix and the desired output signal.

[0011] According to the above technical solution, the predistortion coefficient lookup table is stored in the FPGA register.

[0012] According to the above technical solution, the methods for converting predistortion coefficients into lookup table form include: For each combination of nonlinear order and memory depth in all terms of the predistortion model, a lookup table is created; each lookup table includes multiple storage units. Based on the magnitude range of the original signal, the same number of quantization intervals are evenly divided in each lookup table, and an index address is assigned to each quantization interval so that the index addresses of different quantization intervals in the same lookup table are different, and the index addresses of the same quantization intervals in different lookup tables are the same. The calculated predistortion coefficients are stored in the corresponding storage units of the corresponding lookup table according to their nonlinear order, memory depth and the instantaneous modulus of the corresponding original signal.

[0013] According to the above technical solution, the methods for pre-distortion compensation include: Calculate the average magnitude of the original signal; Based on the combination of predistortion coefficients obtained from the lookup table, the corresponding nonlinear order and memory depth, and the average modulus and the instantaneous modulus, a predistortion signal of the original signal to be processed is generated, and the predistortion signal is input into the gallium nitride power amplifier.

[0014] Another aspect of the present invention provides a digital predistortion system for a gallium nitride power amplifier, comprising: The lookup table construction module is used to collect the original signal before it is input to the gallium nitride power amplifier and the feedback signal output after the original signal is amplified by the gallium nitride power amplifier, and perform synchronization alignment processing; calculate the instantaneous magnitude value corresponding to each sampling point in the synchronized original signal, and perform averaging to obtain the average magnitude value; construct all terms of the generalized memory polynomial model based on the synchronized original signal and the instantaneous magnitude value corresponding to each sampling point; add a moving average generalized memory polynomial term to the generalized memory polynomial model by introducing the average magnitude value to obtain the predistortion model; estimate the coefficients of all terms in the predistortion model based on the synchronized feedback signal, and convert them into a predistortion coefficient lookup table indexed by the instantaneous magnitude value, and pass it to the predistortion compensation module for storage; The predistortion compensation module is used to calculate the instantaneous modulus of the original signal, query the stored predistortion coefficient lookup table based on the instantaneous modulus, obtain the predistortion coefficient, and perform predistortion compensation on the original signal. The power amplifier module is used to amplify and output the pre-distortion compensated signal.

[0015] The beneficial effects of this invention are as follows: Based on a pre-constructed predistortion coefficient lookup table indexed by instantaneous magnitude, the invention obtains predistortion coefficients. After predistortion compensation of the original signal to be processed, the signal is input to a gallium nitride power amplifier for amplification and output. This achieves high-precision compensation for signal distortion while maintaining high processing efficiency. In constructing the predistortion coefficient lookup table: firstly, the instantaneous magnitude of the original signal after synchronization alignment is calculated and averaged to obtain the average magnitude indicating the signal envelope variation characteristics, providing a key parameter for accurately modeling the long memory effect of gallium nitride power amplifiers; then, a generalized memory polynomial model is constructed based on the original signal and its instantaneous magnitude. By introducing the average magnitude, a moving average generalized memory polynomial term is added. Based on the generalized memory polynomial model describing traditional nonlinear characteristics, a predistortion model capable of capturing the long memory effect is further formed, significantly enhancing the modeling accuracy and compensation specificity of gallium nitride power amplifier distortion behavior.

[0016] Furthermore, this invention employs a recursive averaging algorithm based on dual time constants to average the instantaneous magnitude of the original signal, thereby achieving accurate tracking of the dynamic changes in the signal envelope by the average magnitude.

[0017] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a digital predistortion method for a gallium nitride power amplifier according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a digital predistortion system for gallium nitride power amplifiers using existing technology; Figure 3 This is a structural diagram of a digital predistortion system for a gallium nitride power amplifier according to an embodiment of the present invention; Figure 4 This is another structural diagram of a digital predistortion system for a gallium nitride power amplifier according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0021] It should be noted that the illustrations provided in the embodiments of the present invention are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0022] In this invention, it should also be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the terms "first" and "second" are used only for descriptive and distinguishing purposes and should not be construed as indicating or implying relative importance.

[0023] Example 1 This invention provides a digital predistortion method for gallium nitride power amplifiers, the process of which is as follows: Figure 1 As shown, the steps include: S1. Calculate the instantaneous magnitude of the original signal. The instantaneous magnitude is the basis for subsequent predistortion coefficient lookup and magnitude averaging.

[0024] S2. Based on the instantaneous modulus value in step S1, look up the table to obtain the predistortion coefficient and perform predistortion compensation.

[0025] Methods for constructing predistortion coefficient lookup tables include: (1) Acquire the original signal before it enters the gallium nitride power amplifier and the feedback signal after the original signal is amplified by the gallium nitride power amplifier, and perform synchronization alignment processing. Synchronization alignment is to ensure that the original signal and the feedback signal are matched in time, amplitude and phase when the coefficients are calculated in the future, which is a prerequisite for accurately solving the predistortion coefficients.

[0026] Furthermore, the feedback signal is acquired by extracting it through coupling at the output of the gallium nitride power amplifier. This coupling method extracts only a very small portion of the power output signal from the GaN power amplifier as the feedback signal, thus not affecting the signal transmission of the main transmit link while providing effective reference data for DPD coefficient calculation.

[0027] Furthermore, methods for synchronizing and aligning the original signal and the feedback signal include: Based on the acquired raw and feedback signals, a cross-correlation function is constructed, and the maximum cross-correlation function magnitude and its corresponding time offset are found. The calculation length N of the cross-correlation function is typically greater than or equal to 2048. This calculation can accurately locate the time deviation between the raw and feedback signals, solving the asynchrony problem caused by link delay during signal transmission.

[0028] The maximum cross-correlation function magnitude is compared with a preset threshold. If it is greater than the preset threshold, synchronization is considered successful. The time offset corresponding to the maximum cross-correlation function magnitude is used as the time delay between the original signal and the feedback signal, and time delay compensation is performed on the feedback signal based on this time delay. If it is less than or equal to the preset threshold, synchronization is considered unsuccessful, and the original signal and feedback signal are reacquired. Successful synchronization is a prerequisite for subsequent amplitude and phase adjustments and coefficient calculations. Recalculating upon synchronization failure avoids errors in pre-distortion coefficient calculation caused by signal asynchrony, ensuring the DPD compensation effect.

[0029] The amplitude and phase of the feedback signal after time delay compensation are adjusted; wherein, the amplitude adjustment coefficient is related to the ratio of the magnitude of the maximum cross-correlation function to the power of the feedback signal, and the phase adjustment angle is the phase angle of the magnitude of the maximum cross-correlation function.

[0030] (2) Calculate the instantaneous magnitude value corresponding to each sampling point in the original signal after synchronization alignment, and perform averaging to obtain the average magnitude value. The magnitude averaging process is designed to address the long-term memory characteristics caused by the Trapping effect of GaN power amplifiers.

[0031] Furthermore, the method for averaging the instantaneous magnitude values ​​corresponding to each sampling point in the original signal after synchronization alignment includes: For each sampling point, compare its instantaneous modulus with the average modulus corresponding to the previous sampling point. When the instantaneous magnitude of a sampling point is greater than or equal to the average magnitude of the previous sampling point, the average magnitude of that sampling point is adjusted using the charging time constant; when the instantaneous magnitude of a sampling point is less than the average magnitude of the previous sampling point, the average magnitude of that sampling point is adjusted using the discharging time constant. The charging and discharging time constants are inherent parameters of analog electronic devices. This averaging method closely matches the electrical characteristics of GaN power amplifiers, more accurately reflecting the trend of signal magnitude changes, and providing the MA_GMP model with input parameters that conform to the device characteristics.

[0032] (3) Based on the original signal after synchronization alignment and the instantaneous magnitude values ​​corresponding to each sampling point, construct all terms of the generalized memory polynomial model. The generalized memory polynomial (GMP) model includes diagonal terms and cross terms. The diagonal terms are obtained by multiplying the diagonal term coefficients with the nonlinear transformation term of the input signal, and the cross terms are obtained by multiplying the cross term coefficients with the corresponding cross nonlinear transformation term. This model is the core model of traditional DPD technology for LDMOS power amplifiers. Here, the basic GMP terms are constructed first to adapt to GaN power amplifiers.

[0033] Furthermore, the generalized memory polynomial term, based on the synchronized original signal and the instantaneous modulus corresponding to each sampling point, constructs nonlinear transformation terms and cross-nonlinear transformation terms at different memory depths. These transformation terms are then multiplied by their corresponding model coefficients and summed. Memory depth, nonlinear order, cross-term, and cross-term memory depth are the core parameters of the GMP model. Different combinations of these parameters cover the distortion characteristics of power amplifiers with different memory lengths and degrees of nonlinearity.

[0034] (4) By introducing the average modulus, a moving average generalized memory polynomial term is added to the generalized memory polynomial model to obtain the predistortion model. The moving average generalized memory polynomial (MA_GMP) term is a compensation term added to address the long-term memory effect of GaN power amplifiers, making the overall predistortion model a GMP+MA_GMP composite model, which can effectively solve the problem of the traditional GMP model failing to meet the performance standards of GaN power amplifiers.

[0035] Furthermore, the output of the predistortion model includes the sum of generalized memory polynomial terms and moving average generalized memory polynomial terms. The output of the composite model is the superposition of GMP terms (including diagonal and cross terms) and MA_GMP terms (including diagonal and cross terms). Through the complementarity of the two models, the long memory distortion caused by the GaN power amplifier trapping effect is offset, achieving a better linear compensation effect.

[0036] The moving average generalized memory polynomial term is constructed based on the original signal after synchronization alignment and the instantaneous magnitude value corresponding to each sampling point. Nonlinear transformation terms and cross-nonlinear transformation terms at different memory depths are constructed. Each transformation term is multiplied by the corresponding model coefficients, and then multiplied by the average magnitude value. Finally, the terms are summed.

[0037] (5) Based on the feedback signal after synchronization alignment, estimate the coefficients of all terms in the predistortion model and convert them into a predistortion coefficient lookup table indexed by the instantaneous modulus value. The coefficient estimation is the process of solving the model parameters by constructing a linear system. Converting it into a lookup table form can realize fast table lookup compensation in FPGA, meet the needs of real-time processing of RF signals, and adapt to the high-frequency and high-power application scenarios of GaN power amplifiers.

[0038] Furthermore, methods for estimating predistortion coefficients include: Based on all terms of the predistortion model, a regression matrix is ​​constructed. The regression matrix U is an M*N matrix (M is the number of samples, i.e., the length of the original signal x0 after synchronization; N is the number of terms in the GMP+MA_GMP model), and each row corresponds to the nonlinear transformation result of one term in the model, which is the core matrix for solving the coefficients.

[0039] The desired output signal is constructed based on the error between the synchronized original signal and the feedback signal. The desired output signal includes an error term, specifically the error between the synchronized feedback signal and the original signal after compensation by the predistortion model. This signal is constructed to solve for the optimal predistortion coefficients by minimizing the error.

[0040] The predistortion coefficients are obtained by solving the linear system formed by the regression matrix and the desired output signal. Solving the linear system typically involves matrix decomposition methods (such as Cholesky decomposition) to calculate the inverse matrix. This method is computationally efficient, numerically stable, and suitable for the requirements of real-time RF signal processing.

[0041] Furthermore, the predistortion coefficient lookup table is stored in the FPGA's registers. The high-speed read / write capabilities of the FPGA registers can meet the processing requirements of high sampling rates for radio frequency signals.

[0042] Furthermore, methods for converting predistortion coefficients into lookup table form include: For each combination of nonlinear order and memory depth in all terms of the predistortion model, a lookup table is created; each lookup table includes multiple storage units. The number of lookup tables is determined by the maximum number of terms in the model, and each table corresponds to a set of nonlinear order and memory depth combinations.

[0043] Based on the magnitude range of the original signal, the lookup tables are evenly divided into multiple quantization intervals of equal number. An index address is assigned to each quantization interval, ensuring that different quantization intervals within the same lookup table have different index addresses, and the same quantization intervals within different lookup tables have the same index address. The magnitude index corresponds to the quantization interval. Uniform quantization guarantees the continuity of the magnitude index and the accuracy of the lookup, while the design of identical index addresses facilitates synchronous lookup calculations across multiple tables.

[0044] The calculated predistortion coefficients are stored in the corresponding storage units of the corresponding lookup table according to their nonlinear order, memory depth, and the instantaneous magnitude of the corresponding original signal. This storage method enables rapid location of the corresponding coefficients based on the instantaneous magnitude, avoiding the complex calculations of real-time model terms and improving the processing efficiency of predistortion compensation.

[0045] Furthermore, the process of constructing the lookup table is not executed only once during system initialization, but is dynamically triggered and updated according to preset rules to adapt to the drift of long memory characteristics caused by the Trapping effect of GaN power amplifier.

[0046] There are two specific triggering methods: Periodic triggering: Triggered at fixed time intervals to ensure that the predistortion parameters match the real-time characteristics of the power amplifier; Event triggering: Real-time monitoring of the distortion index of the feedback signal to avoid compensation failure caused by sudden changes in power amplifier characteristics.

[0047] The two triggering methods can be enabled independently or used in combination. For example, periodic triggering can be used as the basis, and event triggering can be used as a supplement to balance update efficiency and compensation stability.

[0048] Furthermore, methods for pre-distortion compensation include: Calculate the average magnitude of the original signal; Based on the combination of predistortion coefficients, corresponding nonlinear order, and memory depth obtained from the lookup table, along with the average modulus and the instantaneous modulus, a predistortion signal of the original signal to be processed is generated. This predistortion signal is then input into a gallium nitride power amplifier. The specific predistortion compensation process is as follows: using the instantaneous modulus as an index, the corresponding entries for GMP and MA_GMP are looked up. The entries are multiplied by the corresponding complex signals, and the MA_GMP entry is multiplied by the average modulus. Finally, all results are summed to obtain the predistortion signal. This process can accurately cancel the nonlinear distortion and long memory effect of the GaN power amplifier.

[0049] S3. The signal is amplified and output by the GaN power amplifier. The predistorted signal is converted into an analog small signal by the DAC and then input to the GaN power amplifier. Since the signal characteristics of the predistortion compensation are opposite to the nonlinear distortion characteristics of the GaN power amplifier, the superposition of the two signals achieves linear amplification with low distortion.

[0050] Another aspect of the present invention provides a digital predistortion system for a gallium nitride power amplifier, comprising: The lookup table construction module is used to collect the original signal before it is input to the gallium nitride power amplifier and the feedback signal output after the original signal is amplified by the gallium nitride power amplifier, and perform synchronization alignment processing; calculate the instantaneous magnitude value corresponding to each sampling point in the synchronized original signal, and perform averaging to obtain the average magnitude value; construct all terms of the generalized memory polynomial model based on the synchronized original signal and the instantaneous magnitude value corresponding to each sampling point; add a moving average generalized memory polynomial term to the generalized memory polynomial model by introducing the average magnitude value to obtain the predistortion model; estimate the coefficients of all terms in the predistortion model based on the synchronized feedback signal, and convert them into a predistortion coefficient lookup table indexed by the instantaneous magnitude value, and pass it to the predistortion compensation module for storage; The predistortion compensation module is used to calculate the instantaneous modulus of the original signal, query the stored predistortion coefficient lookup table based on the instantaneous modulus, obtain the predistortion coefficient, and perform predistortion compensation on the original signal. The power amplifier module is used to amplify and output the pre-distortion compensated signal.

[0051] This embodiment provides another digital predistortion system for gallium nitride power amplifiers based on the above method, with the structure as follows: Figure 3 As shown, it includes: Baseband signal data: The raw digital signal that needs to be transmitted.

[0052] DUC: Digital Up-Conversion. It shifts the baseband signal spectrum to a specified radio frequency band.

[0053] CFR: Peak Factor Reduction. This reduces the peak-to-average power ratio of the signal to prevent the signal peak from entering the deep saturation region of the power amplifier, which would cause severe distortion.

[0054] x: The original input signal after DUC / CFR processing. It is the input of the predistorter and also the reference signal for the subsequent synchronization module.

[0055] Predistorter: One of the core modules of this invention. It uses the predistortion coefficients from the "coefficient calculation" module to perform inverse distortion processing on the input signal x to generate the predistorted signal z.

[0056] z: The signal after pre-distortion compensation. This signal itself has "distortion," which is the opposite of the distortion characteristics that the power amplifier will produce.

[0057] DAC: Digital-to-Analog Converter. Converts a digital signal (z) into an analog signal.

[0058] PA: Power amplifier, specifically a gallium nitride (GaN) power amplifier. It is the core device that generates nonlinear distortion and long memory effects. The predistorted signal z is amplified here.

[0059] y A : The amplified analog radio frequency signal output by the power amplifier.

[0060] Coupler: A passive RF device that couples a small portion of energy from the main RF link for feedback monitoring while ensuring that most of the power is transmitted normally.

[0061] RF: Radio Frequency Link, which ultimately transmits signals through an antenna.

[0062] y t The feedback digital signal, obtained through a coupler and converted from analog to digital, represents a sample of the actual output signal of the power amplifier.

[0063] ADC: Analog-to-Digital Converter. It converts the coupled analog feedback signal into a digital signal y. t .

[0064] x0, y0: The original signal and feedback signal after synchronization and alignment. They are signal pairs with precise time, amplitude, and phase matching, used for subsequent coefficient calculations.

[0065] v: Instantaneous magnitude of the original signal.

[0066] A: The average magnitude of the original signal.

[0067] c: The calculated predistortion coefficient.

[0068] The system's process includes: First, the system inputs the original baseband signal to the predistortion compensation module. This module calls the predistortion coefficients latest calculated by the parameter estimation module to perform nonlinear predistortion processing on the signal, generating a predistortion-compensated signal. This signal, after digital-to-analog conversion, is sent to a gallium nitride power amplifier for power amplification. Most of the energy of the amplified RF output signal is transmitted, while a small portion is fed back through a coupler, becoming a feedback signal after analog-to-digital conversion. The signal synchronization and alignment module then acquires the original baseband signal and this feedback signal, performing time and amplitude synchronization. Precise alignment of degree and phase; the aligned two signals, along with the instantaneous and average envelope values ​​obtained from the modulus calculation and processing module, are sent to the parameter estimation module; this module, based on an adaptive model combining generalized memory polynomials and moving average generalized memory polynomials, uses algorithms such as least squares to estimate a new and more accurate set of predistortion coefficients; finally, these updated coefficients are loaded back into the predistortion compensation module in real time, thus forming a continuously optimized feedback control loop that dynamically and accurately compensates for the nonlinear distortion and long memory effect of the gallium nitride power amplifier.

[0069] The methods by which each module performs its function have been described in the method of this embodiment and will not be repeated here.

[0070] Example 2 Based on Example 1, this example provides another digital predistortion method for gallium nitride power amplifiers, including the following steps: T1. The digital signal x undergoes DUC (digital up-conversion) and CFR (peak factor reduction), followed by initial DPD (predistortion compensation). The compensated data z is then converted into a small analog signal via DAC (digital-to-analog converter), and finally transmitted via RF (radio frequency) link after PA (power amplification). At the same time t, partial data x is collected. t and PA's output signal y A Sampling y on the feedback link t To synchronize.

[0071] Specifically, before performing the first predistortion compensation, the predistortion coefficients calculated in the previous calculation without predistortion compensation are used as the initial values ​​of the predistortion coefficients and stored in the predistorter. The predistortion coefficients are then updated and iterated through subsequent calculations.

[0072] T11. Data acquisition is triggered by a preset period or event, and a portion of the data x of the digital signal x is acquired simultaneously at time t. t and PA's output signal y A Sampling y on the feedback link t .

[0073] T12, regarding x t and yt Perform synchronization alignment: construct x t and y t The cross-correlation function is expressed as follows:

[0074] Where n is the index of the discrete sampling point, used to traverse the signal x t and y t Each sample point is used to calculate the cross-correlation value; N is the correlation calculation length, which can be 2048, 4096 or other values, and is usually greater than or equal to 2048; Representing complex data . conjugate.

[0075] T13. Find the time t with the maximum correlation modulus. max Record the maximum correlation value: maxCorr = max(abs(corr(t)) max ))).

[0076] Compare maxCorr with the preset threshold.

[0077] If maxCorr > threshold, then synchronization is successful, and the delay is taken as t. max Otherwise, synchronization fails, the calculation ends, the predistortion coefficients are not updated, and the original predistortion coefficients remain unchanged.

[0078] If synchronization is successful, then use the delay t. max For y t Compensation is performed to obtain the compensated data x0=x t y0(t)=y t (t+t max ).

[0079] T14. Adjust the amplitude and phase of the feedback signal.

[0080] The power value p of the feedback data is calculated using the following formula:

[0081] Based on the power value p, the amplitude and phase of the compensation feedback signal are calculated as follows: y0(t) = y0(t) * corr(t) max ) / p.

[0082] T2. Find the instantaneous magnitude v of the digital signal x and perform averaging.

[0083] T21. The formula for calculating the modulus of a digital signal x is as follows:

[0084] in, Indicates signal The modulus; Indicates signal The real part; Indicates signal The imaginary part.

[0085] T22, Perform modulus averaging.

[0086] Let the initial plane modulus value A(0) = v(0), and update the value of A(n) according to the following formula:

[0087] in, For data sampling rate, and To simulate the charging and discharging time constants of electronic devices, Let x be the magnitude of the digital signal.

[0088] Furthermore, although the initial plane magnitude A(0) can be set as the magnitude of the first sampled data, its value is arbitrary from the perspective of practical engineering applications. The average calculation of the magnitude is achieved by a composite filtering mechanism based on two alpha filters (corresponding to the charging time constant and discharging time constant of the analog electronic device, respectively). This filtering mechanism has dynamic smoothing characteristics and can operate normally regardless of the initial state of the system. After startup, only a short sampling point (about 1000) is needed for smoothing, and the average magnitude will quickly converge to a stable and uniform value, completely eliminating the influence of the initial value on the final average result.

[0089] T3. Calculate the DPD coefficients using the synchronized data x0 and y0.

[0090] T31. Calculate the expected output including the error term using the synchronized data x0 and y0. The calculation formula is as follows:

[0091] in, The synchronized digital signal x, The feedback signal y after synchronization This is the error term.

[0092] T32. Construct U using x0, where U is an M*N matrix, M is the number of samples (i.e., the length of x0), and N is the number of terms in the GMP+MA_GMP model. Each term in the model contains different values ​​of m, k, and q. m is the memory depth, representing the historical delay of the signal; k is the power of the amplitude, representing the nonlinearity of the signal amplitude; and q is a phase or delay-related parameter, related to the phase modulation or additional delay of the signal. Each row in matrix U corresponds to a term in the GMP or MA_GMP model, and the term has the form: .

[0093] Specifically, the expression for the GMP model is:

[0094] in, These are the diagonal terms in the GMP model; For the cross term in the GMP model; z(n): the output signal of PD at time n; x(nq): the input signal of PD at time nq, where q is the memory depth term; a kq : Diagonal coefficients in GMP; b klm : Cross term coefficients in GMP; k: nonlinear order; q: memory depth in the memory polynomial; m: cross degree; l: memory depth in the cross term; K: Maximum order of the cross term; K: Maximum order of the diagonal term; Q: Maximum memory depth of the cross term; Q: Maximum memory depth of the diagonal term; M: Maximum memory depth of the cross term; b : Maximum intersection depth.

[0095] The expression for the GMP+MA_GMP model is:

[0096] in, These are the diagonal terms in the GMP model; These are the cross terms in the GMP model; For the diagonal terms in the MA_GMP model; For the cross term in the MA_GMP model; z(n): the output signal of PD at time n; x(nq): the input signal of PD at time nq, where q is the memory depth term; a kq : Diagonal coefficients in the GMP model; b klm : Cross term coefficients in the GMP model; c kq : Diagonal coefficients in the MA_GMP model; d klm : Cross term coefficients in the MA_GMP model; k: nonlinear order; q: memory depth in the memory polynomial; m: cross degree; l: memory depth of the cross term; v: moving average.

[0097] Perform linear computation on matrix U: (1) Calculation H represents the conjugate transpose of the matrix.

[0098] (2) Calculation .

[0099] (3) Find the matrix inverse matrix In this embodiment, a matrix decomposition method, such as Cholesky decomposition, is used.

[0100] (4) Calculate the coefficients A represents different groups of a corresponding to different instantaneous moduli. kq b klm c kq d klm A collection of things.

[0101] T33. Convert coefficients into a lookup table: , To and The corresponding model parameters are: abs is the modulus index, which takes the value 0~511 in this embodiment, and the signal modulus corresponding to the abs value is abs*64; each LUT (Look-Up Table) has a length of 512, and the number of tables is determined by the maximum number of terms in the model used, usually 24; n is the table index.

[0102] Update the lookup table and write the obtained LUT into the FPGA registers.

[0103] T4. Perform final predistortion compensation on the input signal x based on the predistortion coefficient c and average modulus A obtained from the table lookup.

[0104] T41. Using the input signal at time t and the magnitude of the input signal within the memory depth range before the current time, floor(v(tm) / 64), as the index, look up... and The lookup result of the nth table corresponding to the modulus value at time tm is: , where 0≤m≤M, and M is the maximum memory depth. This is the table entry corresponding to the GMP model. The modulus value in the table entry corresponds to time tm, the signal time is tq, and the table index is n. This is the table entry corresponding to the MA_GMP model. The magnitude value in the table entry corresponds to time tm, and the signal time is tq.

[0105] T42. The lookup table results are multiplied one by one with the corresponding input signal data. The time of the complex signal data corresponding to the lookup table result is determined by a term in the adopted GMP+MA_GMP model. The multiplication result is... .

[0106] T43. Multiply the terms in the MA_GMP model by the average modulus. The result of the multiplication is: .

[0107] T44. Sum all the results to obtain the compensated output:

[0108] The compensation output is then input into a gallium nitride power amplifier for amplification.

[0109] This embodiment also provides a digital predistortion system for a gallium nitride power amplifier, the structure of which is as follows: Figure 4 As shown, it includes: Baseband signal data: The raw baseband digital signal of the communication system, which is the input source of the entire system.

[0110] DUC / CFR: DUC (Digital Upconverter): Upconverts baseband signals to intermediate frequency / radio frequency bands, while simultaneously increasing the sampling rate through interpolation filtering, and outputting digital signals suitable for subsequent processing.

[0111] CFR (Peak Factor Reduction): Reduces the peak-to-average power ratio (PAPR) of the signal, decreasing the probability of the GaN PA entering the deep nonlinear region due to high peak signals, and suppressing out-of-band spurious signals. The output signal is denoted as x.

[0112] Predistorter: The core execution unit of the DPD system. Based on the predistortion coefficients c provided by the coefficient calculation module, it applies inverse distortion to the input signal x to generate the signal z. This inverse distortion cancels out the nonlinear distortion of the GaN PA, thus linearizing the final output.

[0113] DAC (Digital-to-Analog Converter): Converts the digital predistorted signal z into an analog signal and sends it to the GaN PA.

[0114] PA (Gallium Nitride Power Amplifier): Amplifies analog signals to output high-power radio frequency signals. A .

[0115] RF (Radio Frequency): Processed high-power radio frequency signals are radiated through an antenna to achieve wireless communication.

[0116] Coupler: The radio frequency signal y output from PA A A small portion of the power is coupled out for feedback detection.

[0117] ADC (Analog-to-Digital Converter): Converts coupled analog radio frequency signals into digital signals. t , and send it to the synchronization module.

[0118] Synchronization module: Receives the reference signal x from the transmit link. t (Delayed version of x) and feedback y t Synchronization alignment of time, amplitude, and phase is performed.

[0119] Modulus Calculation Module: Calculates the modulus v (i.e. signal amplitude) of the input signal x0 after synchronization, which is used to characterize the intensity characteristics of the input signal.

[0120] Average module: Performs a moving average on the modulus v and outputs A.

[0121] Coefficient calculation module: Based on the synchronized x0, y0 and average modulus A, combined with the GMP (Generalized Memory Polynomial) + MA_GMP (Moving Average Generalized Memory Polynomial) model, the predistortion coefficient c is solved by matrix operations (such as the least squares method).

[0122] The process executed by this system includes: The operation of this digital predistortion system begins with baseband signal data, which first undergoes upconversion and peak-to-average ratio optimization preprocessing via DUC (digital upconversion) and CFR (peak factor reduction) modules. Subsequently, the preprocessed signal enters the core predistorter module, which applies a predistortion correction to the signal with the opposite effect to the nonlinear distortion characteristics of the gallium nitride power amplifier by using the latest predistortion coefficients dynamically loaded from the coefficient calculation module. The predistorted signal is then converted into an analog signal by a DAC (digital-to-analog converter) and input to the gallium nitride power amplifier for power amplification. Most of the energy of the amplified RF signal is transmitted through the RF link, while a small portion of the energy is coupled to the feedback path via a coupler and sampled as a digital feedback signal by an ADC (analog-to-digital converter).

[0123] The system synchronously acquires the original signal from the transmission path and the signal from the feedback path, and performs precise time, amplitude, and phase alignment in the synchronization module. The aligned original signal is sent to the modulus calculation module to calculate its instantaneous modulus value, and then the averaging module processes it using a unique recursive averaging algorithm to generate an average modulus value that can characterize the historical changes in the signal envelope. This is intended to accurately capture the long memory effect of the gallium nitride power amplifier.

[0124] The coefficient calculation module integrates the aligned original signal, feedback signal, instantaneous modulus and average modulus, and performs parameter estimation based on an enhanced model that combines generalized memory polynomial and moving average generalized memory polynomial. This results in a new generation of more accurate predistortion coefficients, which are then immediately updated in the predistorter module.

[0125] Thus, the system forms a complete adaptive closed loop: the forward path continuously performs real-time pre-distortion compensation on the signal, while the feedback path continuously monitors the power amplifier output and dynamically optimizes the pre-distortion parameters, thereby effectively suppressing the nonlinear distortion of the gallium nitride power amplifier, especially the long memory effect.

[0126] In summary, this invention provides a digital predistortion method and system for gallium nitride power amplifiers, which achieves high-precision compensation for signal distortion of gallium nitride power amplifiers while having high processing efficiency.

[0127] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0128] The order of the steps in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0129] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A digital predistortion method for gallium nitride power amplifiers, characterized in that, include: The instantaneous magnitude of the original signal is calculated, the predistortion coefficient is obtained by looking up the table based on the instantaneous magnitude, predistortion compensation is performed, and then the signal is amplified and output by a gallium nitride power amplifier. The methods for constructing the predistortion coefficient lookup table include: The original signal before input to the gallium nitride power amplifier and the feedback signal output after amplification by the gallium nitride power amplifier are acquired and synchronized. The instantaneous magnitude value corresponding to each sampling point in the original signal after synchronization and alignment is calculated and averaged to obtain the average magnitude value. Based on the original signal after synchronization alignment and the instantaneous modulus values ​​corresponding to each sampling point, all terms of the generalized memory polynomial model are constructed. By introducing the average modulus value, a moving average generalized memory polynomial term is added to the generalized memory polynomial model to obtain the predistortion model. Based on the feedback signal after synchronization alignment, the coefficients of all terms in the predistortion model are estimated and converted into a predistortion coefficient lookup table indexed by the instantaneous modulus value.

2. The digital predistortion method for gallium nitride power amplifiers according to claim 1, characterized in that, The feedback signal is acquired by extracting it through coupling at the output of the gallium nitride power amplifier.

3. The digital predistortion method for gallium nitride power amplifiers according to claim 1, characterized in that, Methods for synchronizing and aligning the original signal and the feedback signal include: Based on the acquired raw and feedback signals, a cross-correlation function is constructed, and the maximum cross-correlation function modulus and its corresponding time offset are found. The maximum cross-correlation function magnitude is compared with a preset threshold. If it is greater than the preset threshold, the synchronization is considered successful. The time offset corresponding to the maximum cross-correlation function magnitude is used as the time delay between the original signal and the feedback signal. The feedback signal is compensated for the time delay based on the time delay. If it is less than or equal to the preset threshold, the synchronization is considered unsuccessful. The original signal and the feedback signal are reacquired. The amplitude and phase of the feedback signal after time delay compensation are adjusted; wherein, the amplitude adjustment coefficient is related to the ratio of the magnitude of the maximum cross-correlation function to the power of the feedback signal, and the phase adjustment angle is the phase angle of the magnitude of the maximum cross-correlation function.

4. The digital predistortion method for gallium nitride power amplifiers according to claim 1, characterized in that, Methods for averaging the instantaneous magnitude values ​​corresponding to each sampling point in the original signal after synchronization and alignment include: For each sampling point, compare its instantaneous modulus with the average modulus corresponding to the previous sampling point. When the instantaneous modulus of the sampling point is greater than or equal to the average modulus of the previous sampling point, the average modulus of the sampling point is adjusted by the charging time constant; when the instantaneous modulus of the sampling point is less than the average modulus of the previous sampling point, the average modulus of the sampling point is adjusted by the discharging time constant.

5. The digital predistortion method for gallium nitride power amplifiers according to claim 1, characterized in that, The output of the predistortion model includes the sum of generalized memory polynomial terms and moving average generalized memory polynomial terms; wherein... The generalized memory polynomial term is constructed based on the original signal after synchronization alignment and the instantaneous modulus value corresponding to each sampling point. Nonlinear transformation terms and cross-nonlinear transformation terms under different memory depths are constructed, and then each transformation term is multiplied by the corresponding model coefficients and summed. The moving average generalized memory polynomial term is constructed based on the original signal after synchronization and alignment and the instantaneous magnitude value corresponding to each sampling point. Nonlinear transformation terms and cross-nonlinear transformation terms at different memory depths are constructed. Each transformation term is multiplied by the corresponding model coefficients, and then multiplied by the average magnitude value. Finally, the terms are summed.

6. The digital predistortion method for gallium nitride power amplifiers according to claim 1, characterized in that, The method for estimating the predistortion coefficients includes: Construct a regression matrix based on all terms of the predistortion model; The desired output signal is constructed based on the error between the original signal after synchronization and the feedback signal; The predistortion coefficients are obtained by solving the linear system formed by the regression matrix and the desired output signal.

7. The digital predistortion method for gallium nitride power amplifiers according to claim 1, characterized in that, The predistortion coefficient lookup table is stored in the FPGA register.

8. The digital predistortion method for gallium nitride power amplifiers according to claim 7, characterized in that, Methods for converting predistortion coefficients into lookup table form include: For each combination of nonlinear order and memory depth in all terms of the predistortion model, a lookup table is created; each lookup table includes multiple storage units. Based on the magnitude range of the original signal, the same number of quantization intervals are evenly divided in each lookup table, and an index address is assigned to each quantization interval so that the index addresses of different quantization intervals in the same lookup table are different, and the index addresses of the same quantization intervals in different lookup tables are the same. The calculated predistortion coefficients are stored in the corresponding storage units of the corresponding lookup table according to their nonlinear order, memory depth and the instantaneous modulus of the corresponding original signal.

9. The digital predistortion method for gallium nitride power amplifiers according to claim 8, characterized in that, Methods for pre-distortion compensation include: Calculate the average magnitude of the original signal; Based on the combination of predistortion coefficients obtained from the lookup table, the corresponding nonlinear order and memory depth, and the average modulus and the instantaneous modulus, a predistortion signal of the original signal to be processed is generated, and the predistortion signal is input into the gallium nitride power amplifier.

10. A digital predistortion system for gallium nitride power amplifiers, characterized in that, include: The lookup table construction module is used to collect the original signal before it is input to the gallium nitride power amplifier and the feedback signal output after the original signal is amplified by the gallium nitride power amplifier, and to perform synchronization alignment processing. Calculate the instantaneous modulus value corresponding to each sampling point in the original signal after synchronization and alignment, and perform averaging to obtain the average modulus value; Based on the original signal after synchronization alignment and the instantaneous modulus value corresponding to each sampling point, all terms of the generalized memory polynomial model are constructed; by introducing the average modulus value, a moving average generalized memory polynomial term is added to the generalized memory polynomial model to obtain the predistortion model; based on the feedback signal after synchronization alignment, the coefficients of all terms in the predistortion model are estimated and converted into a predistortion coefficient lookup table indexed by the instantaneous modulus value, which is then passed to the predistortion compensation module for storage. The predistortion compensation module is used to calculate the instantaneous modulus of the original signal, query the stored predistortion coefficient lookup table based on the instantaneous modulus, obtain the predistortion coefficient, and perform predistortion compensation on the original signal. The power amplifier module is used to amplify and output the pre-distortion compensated signal.

Images