Digital predistortion method and system based on baseband-RF joint optimization
Through the digital predistortion method of baseband-RF joint optimization, the model coefficients are updated using the recursive least squares algorithm, which solves the problem of insufficient out-of-band leakage and linearization performance in the existing technology, and realizes efficient linear optimization of RF power amplifiers.
Patent Information
- Application Number
- CN202111219171.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-20
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2041-10-20
AI Technical Summary
Existing digital predistortion technology cannot effectively optimize the performance of out-of-band leakage, and cannot update linearization performance in real time according to changes in different scenarios, resulting in large errors and poor generalization.
Using a digital predistortion method based on baseband-RF joint optimization, the model coefficients are trained by collecting the data set of the RF power amplifier, combining the baseband excitation signal for frequency conversion processing, calculate the nonlinear intermodulation distortion determining factor, and iterate using the recursive least squares algorithm Update the model coefficients to achieve linear optimization of RF power amplifiers.
Accurate optimization of specific nonlinear distortions is achieved, out-of-band components such as IMD3 are reduced, linearization performance of RF power amplifiers is improved, and linearization requirements of actual systems are met.
Smart Images

Figure CN113949350B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a technology in the field of wireless communications, specifically a digital predistortion based on baseband-RF joint optimization, which can be used in various systems with baseband signal processing and RF power amplifier equipment, including but not limited to LTE, NR and other wireless transmission systems, and can be used in narrowband or broadband signal scenarios. Background Art
[0002] When the power amplifier (RF power amplifier) works in the working range close to the saturation point, nonlinear distortion will be generated. Nonlinearity will cause spectrum regeneration (expansion), adjacent channel interference, in-band distortion, and reduced bit error rate (BER) performance. In order to ensure the linear performance of the RF power amplifier and maintain high efficiency characteristics, the signal distortion effect at the input of the RF power amplifier is compensated, that is, pre-distortion technology. Typical solutions include feedback, feedforward, analog pre-distortion (APD) and digital pre-distortion (DPD), in which APD uses discrete or integrated circuits with analog components to simulate nonlinear distortion effects. Corresponding analog pre-distortion circuits are designed for nonlinear intermodulation distortions (IMDs) such as third-order and fifth-order, and are eliminated by the pre-distortion circuit as the target of RF power amplifier linearization. However, due to complex circuits and / or insufficient linearization, it is difficult to use RF linearization technology as a complete transmitter for base stations. DPD uses an iterative algorithm to obtain output characteristics that are opposite to those of the PA for the baseband mean square error (MSE). However, the algorithm iteration reduces each order of nonlinear distortion on average. However, most high-order nonlinearities and harmonics can be filtered out by bandpass filters, so a DPD solution that specifically optimizes specific nonlinear distortions is needed. Summary of the invention
[0003] In view of the problems that different existing pre-distortion schemes cannot meet the performance requirements of out-of-band leakage, the technology cannot obtain and evaluate the out-of-band distortion at the baseband, the linearization performance cannot change according to different scenarios, and the generalization is poor, the present invention proposes a digital pre-distortion method and system based on baseband-RF joint optimization, and designs a mathematical expression of the out-of-band distortion IMDs including the pre-distortion model coefficients at the baseband, so that the out-of-band components such as IMD3 can be targetedly reduced under the premise of DPD optimizing MSE, thereby realizing digital pre-distortion of baseband-RF joint optimization.
[0004] The present invention is achieved through the following technical solutions:
[0005] The invention relates to a digital predistortion method based on baseband-radio frequency joint optimization. The method comprises the following steps: collecting a data set of transceiver signals of a real radio frequency power amplifier for training a power amplifier model to obtain a model coefficient vector; collecting a data set of baseband excitation signals for inputting into a digital predistortion model, performing up-conversion processing, and then inputting into the power amplifier model to obtain a corresponding radio frequency output signal and a power spectrum diagram thereof; down-converting and linear filtering the radio frequency output signal to obtain a baseband recovery signal, and then obtaining a feedback signal through a post-distortion model to calculate an error and MSE between the excitation and the feedback; and calculating a mathematical expression of a determining factor of nonlinear intermodulation distortion according to the digital predistortion model, the model coefficient vector of the power amplifier model, and up-down conversion related parameters; and finally, deriving a recursive least squares (RLS) improved algorithm according to a joint optimization objective function based on baseband-radio frequency characteristics, and iteratively updating the digital predistortion model coefficient vector according to the algorithm, so that the joint optimization objective function converges, thereby realizing linear optimization of the radio frequency power amplifier, and reducing both the digital index MSE and the radio frequency index IMDs.
[0006] Technical Effects
[0007] The present invention solves the problem that the goal of the optimization algorithm in the existing digital pre-distortion technology is the mean square error. After convergence, it will evenly reduce the nonlinear distortion of each order in and out of the band. In actual situations, only the corresponding distortion that cannot be filtered out needs to be optimized, so it is impossible to achieve high linearization performance. The nonlinear characteristics of the existing analog pre-distortion calculation of the RF power amplifier are relatively vague, so there will be large errors in the compensation through pre-distortion. Moreover, the pre-distortion circuit implemented by circuit design is difficult to update according to changes in device characteristics, and has poor real-time performance.
[0008] Compared with the prior art, the present invention models the frequency domain intermodulation distortion of the RF signal through the coefficients and parameters of different models obtained at the baseband, thereby optimizing the RF index at the baseband and realizing baseband-RF joint optimization. Since some distortions can be removed by bandpass filtering, the improved adaptive algorithm can use the adjustment of the joint objective function to artificially optimize specific nonlinear distortions, such as IMD3, more accurately. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] Figure 1 It is a schematic diagram of the system of the present invention;
[0010] Figure 2 A schematic diagram of the processing process of the digital predistorter module for obtaining a baseband signal;
[0011] Figure 3 Schematic diagram of the process of constructing the iterative module of the adaptive algorithm for the objective function;
[0012] Figure 4 It is a flow chart of the present invention;
[0013] Figure 5 It is a schematic diagram of the effect of the embodiment. DETAILED DESCRIPTION
[0014] like Figure 1 As shown, this embodiment relates to a digital predistortion device based on baseband-RF joint optimization, including: a baseband signal acquisition digital predistorter module, a RF front-end module, a post-distorter feedback signal generation module and an objective function construction adaptive algorithm iteration module, wherein: the baseband signal acquisition digital predistorter module constructs a digital predistortion model according to the time domain data set of the excitation signal of the baseband transmitting end and the DPD coefficient vector from the objective function construction adaptive algorithm iteration module, which is used to add predistortion characteristics to the excitation signal, and then output the processed excitation signal to the RF front-end module and the objective function construction adaptive algorithm iteration module respectively according to the indirect learning structure; the RF front-end module converts the baseband predistortion signal to the RF segment, amplifies the signal power and outputs the signal to the post-distorter feedback signal generation module while transmitting through the antenna ; The post-distorter feedback signal generation module samples and down-converts the output signal to generate a feedback signal data set, adds the same pre-distortion characteristics as DPD to it through the post-distorter, and outputs it to the objective function to build an adaptive algorithm iteration module; The objective function builds an adaptive algorithm iteration module to obtain the baseband signal of the digital pre-distorter module from the baseband signal and the feedback signal from the post-distorter feedback signal generation module to calculate the system error and the digital indicator MSE, and builds the end-to-end input and output model of the overall link of the transmitter according to the vector of the digital pre-distortion model and the vector of the digital up-conversion and power amplification model to calculate the out-of-band intermodulation distortion IMDs expression, thereby constructing a joint optimization objective function of MSE+IMDs, and converge the joint target by improving the RLS adaptive algorithm, and finally achieve the joint optimization of in-band and out-of-band distortion.
[0015] The RF front-end module includes: up-converter, down-converter, A / D converter and real RF power amplifier, wherein: the baseband signal is converted to the RF section by the up-converter and DAC and then outputted. RF (n) = v (cosω1n + cosω2n) to the real RF power amplifier, which generates system nonlinear distortion while performing power amplification. The real RF power amplifier outputs the RF signal y RF (n) The signal is transmitted through the antenna and output to the post-distortion feedback signal generation module, and then processed by the ADC, down-converted to baseband, and processed by the post-distortion to generate a feedback signal, such as Figure 3 As shown, Among them: The third-order intermodulation distortion (IMD3) generated by the RF power amplifier is the frequency component of the frequency point (2ω1±ω2) or (2ω2±ω1), and its size is determined by the factor It is determined that the purpose of reducing IMD3 can be achieved by reducing the model coefficient b3, ω1 and ω2 are dual-tone points, n is the data set sample point, and v is the signal amplitude.
[0016] According to the above analysis, it is deduced that the factors that determine the size of out-of-band distortion such as IMD3 are the odd-order coefficients in the polynomial. Therefore, a new IMDs mathematical expression is designed by obtaining the corresponding coefficients of the system end-to-end (DPD-PA) model.
[0017] In this embodiment, in order to calculate the mathematical expression of RF IMDs, a data set of RF signals transmitted and received at both ends of a real RF power amplifier is sampled and obtained during the initialization phase. The power amplifier model and its coefficient vector b are trained using the RLS algorithm and the sampled data. The model is then transmitted to the objective function through the post-distortion filter feedback signal generation module to construct an adaptive algorithm iteration module, and no subsequent updates are required.
[0018] The baseband signal acquisition digital predistorter module comprises: a baseband signal acquisition unit and a digital predistorter generation and update unit, wherein: the baseband signal acquisition unit is connected to the baseband physical layer module and receives the OFDM signal generated by it, the digital predistorter generation and update unit generates an initialized DPD model, and then updates the DPD model according to the DPD model coefficients iterated by the adaptive algorithm iteration module constructed according to each objective function, until the joint optimization target meets the required threshold, at which time the final DPD model is generated and applied to the transceiver system to obtain a baseband signal result containing predistortion.
[0019] The post-distorter feedback signal generating module comprises: a down-conversion unit and a post-distorter unit, wherein: the down-conversion unit is connected to the RF front-end module and samples and transmits RF transmission signal information, and down-converts the RF signal to baseband; the post-distorter unit performs post-distortion adding processing according to the transmission signal information after down-conversion; the post-distorter is a pre-distortion model having the same input and output characteristics as the digital pre-distortion to obtain the feedback signal data set result.
[0020] The objective function construction adaptive algorithm iteration module includes: a system error and digital indicator calculation unit, an analog indicator IMD3 mathematical expression calculation unit and an improved RLS adaptive algorithm iteration unit, wherein: the system error and digital indicator calculation unit is connected to the baseband signal acquisition digital predistorter module and the post-distorter feedback signal generation module and obtains the baseband predistortion signal and the feedback predistortion signal, the difference between the two is the system nonlinear error, and the digital indicator MSE is calculated based on this; the analog indicator IMD mathematical expression calculation unit performs IMD3 expression calculation according to the current DPD model coefficient, the up-conversion modulation formula and the PA model coefficient to obtain the equivalent formula of the analog indicator IMD; the improved RLS adaptive algorithm iteration unit jointly optimizes the objective function by jointly designing the digital / analog indicators, derives the model coefficient recursive formula of the improved adaptive algorithm, performs iterative update calculation of the DPD model coefficient, and finally obtains the coefficient least squares solution that makes the objective function converge to the specified threshold.
[0021] The joint linearization of in-band and out-of-band distortions refers to: calculating the IMDs mathematical expression as an analog indicator, thereby constructing a joint optimization objective function based on baseband-RF characteristics based on the digital / analog indicator, and iterating the digital pre-distortion model coefficient vector a using a new improved RLS adaptive algorithm derived from the joint optimization objective function, and outputting it to the baseband signal to obtain the digital pre-distortion module to update the digital pre-distortion model in real time. When the joint indicator converges, the iterative update is stopped. At this time, the generated digital pre-distortion model can realize the joint linearization of the system's in-band and out-of-band distortions.
[0022] The power amplification model is: RF (n) = b1z RF (n)+b2z RF (n) 2 +b3z RF (n) 3 , where: b = [b1, b2, b3] is the model coefficient vector, y RF (n) is the output RF signal.
[0023] The digital pre-distortion model is: z(n)=a1x(n)+a2x(n) 2 +a3x(n) 3 , where: a = [a1, a2, a3] is the pre-distorter model coefficient, x(n) is the baseband input signal; the memoryless model of the post-distorter: Where: y(n) is the baseband output signal.
[0024] The end-to-end input and output model of the transmitter overall link is: RF (n) = b1cosωc [a1x(n)+a2x(n) 2 +a3x(n) 3 ]+b2cosω c 2 [a1x(n)+a2x(n) 2 +a3x(n) 3 ] 2 +b3cosω c 3 [a1x(n)+a2x(n) 2 +a3x(n) 3 ] 3 =c1x(n)+c2x(n) 2 +c3x(n) 3 …+c9x(n) 9 , where: x(n) is the baseband input signal, and the coefficients of the odd terms are: c3=a3b1 cosω c +2a1a2b2cosω c 2 +a1 3 b3cosω c 3 , c5=2a2a3b2cosω c 2 +(3a1 2 a3+a1a2 3 )b3cosω c 3 , c7=(3a1a3 2 +3a2 2 a3)b3cosω c 3 .
[0025] The joint optimization objective function based on baseband-RF characteristics is: Obj=MSE+IMDs, where: IMDs=αc3+βc5+γc7, α, β, γ are weight coefficients respectively; Where N is the sample length of the data set, and since the predistorter and postdistorter provide the same input and output characteristics, y(n)-x(n) can be equivalent to the error
[0026] like Figure 4 As shown, this embodiment relates to a digital predistortion method based on baseband-RF joint optimization of the above system, including:
[0027] Step 1: Based on the changes in the baseband signal, evaluate whether the end-to-end system needs to add a new digital pre-distortion model, or whether the trained digital pre-distortion model is applicable to the current excitation signal. Make a judgment based on user needs and baseband signal configuration. The specific judgment criteria and methods are: observe the AM-AM characteristics of the PA transceiver, the NMSE performance of the baseband transceiver data set, and the out-of-band leakage of the RF signal frequency domain.
[0028] Step 2: Initialize all parameters in the adaptive algorithm, including the forgetting factor λ, regularization parameter δ, inverse correlation matrix P(n), and IMDs expression control factors α, β, and γ used to control the stability of the system, the convergence speed and accuracy of the algorithm; initialize the digital pre-distortion model coefficient vector a(0) so that the digital pre-distortion model provides the simplest linear change for the excitation signal in the initial state; the baseband signal is converted to the RF section after passing through the DPD initial model and output to the real RF power amplifier. The corresponding sampling frequency is configured according to the bandwidth of the baseband signal and the carrier frequency of the up and down conversion, and the RF signals sent and received at both ends of the PA are sampled to train the power amplifier model and obtain the model coefficient vector b.
[0029] The training described uses an RLS adaptive algorithm to train the model coefficient vector so that the model fits the input and output characteristics of a real RF power amplifier.
[0030] Step 3: Update the digital pre-distortion model according to a(n), input the baseband signal into the digital pre-distortion model, and then pass it through the DAC, up-convert it to the RF section, and then output it to the real RF power amplifier to increase the signal transmission power and generate nonlinear distortion.
[0031] Step 4: Down-convert the PA output RF signal, obtain the baseband output signal from the ADC and transmit it to the post-distorter. The post-distorter generates a feedback signal for indicator calculation and algorithm iteration.
[0032] Step 5: According to the coefficient vector of the digital pre-distortion model and the power amplifier model and the digital up-conversion characteristics, the expressions of the odd-order coefficients of the end-to-end model are calculated, the mathematical expression of the IMDs control factor is derived, and the system error and the mean square error MSE are calculated according to the feedback signal and the excitation signal. The MSE and IMDs indicators are used to construct a jointly optimized MSE+IMDs objective function, and the objective function is gradually iterated and converged to the least squares solution through the derived improved recursive least squares adaptive algorithm, which specifically includes:
[0033] Step 5.1: Calculate the gain vector matrix k(n) according to the inverse correlation matrix, the forgetting factor, and the feedback signal to control the iterative update variation of the digital predistortion model coefficient vector.
[0034] Step 5.2: Calculate the system objective function and update the expressions of system error e(n) and IMDs control factor.
[0035] Step 5.3, iteratively update the digital predistortion model coefficient vector a(n) according to the gain vector, forgetting factor, inverse correlation matrix, system error, IMDs and their control factors.
[0036] Step 5.4: Update the inverse correlation matrix.
[0037] Step 5.5: Determine whether the joint optimization index has converged. If not, repeat steps 3 to 5.5.
[0038] Step 6: Apply the converged digital pre-distortion model to the transceiver system, linearize the excitation signal, calculate the linearity index to evaluate the algorithm performance, and measure the RF signal spectrum of the RF power amplifier output to observe the out-of-band distortion.
[0039] When this embodiment is applied in a dual-tone signal scenario, when the distance between the dual-tone audio points is appropriate, IMD3 is the out-of-band distortion with the greatest impact. Since the out-of-band leakage characteristics of the dual-tone signal are simple, the power values of each order of intermodulation distortion frequency points can be directly observed on the spectrum diagram. Therefore, the linearization performance of the DPD joint optimization algorithm is evaluated by calculating the baseband normalized mean square error NMSE of the input and output signals of the RF power amplifier and the spectrum IMD3, where IMD3 is defined as the ratio of the third-order intermodulation frequency power to the baseband power.
[0040] In this scenario, two baseband dual-tone signals with different frequencies are used for testing. The dual-tone signal frequency is 200 / 250 Hz or 150 / 600 Hz, the carrier frequency is 0.1 MHz, the sampling frequency is 20 or 30 MHz, and the sampling length is 400,000 or 600,000.
[0041] In this embodiment, the mean square error and third-order intermodulation distortion performance results of the dual-tone signal scenario are obtained by using a dual-tone signal as shown in Table 1. As can be seen from Table 1, the joint algorithm has improved the mean square error, third-order intermodulation and fifth-order intermodulation indicators compared with the existing DPD. The joint optimization DPD solution for IMD3 can collect 5-6dB of IMD3 gain, and the joint optimization algorithm for IMD5 can also obtain 4-6dB of IMD5 gain by adjusting the control factor. However, the targeted optimization of a certain indicator will increase other nonlinear distortions, so a reasonable trade-off needs to be made according to the needs of the user.
[0042] Table 1 Performance indicators of various methods in dual-tone scenarios
[0043]
[0044] In this embodiment, in the broadband signal scenario, due to its complex spectral characteristics, it is impossible to directly find the third-order intermodulation frequency point to measure its out-of-band distortion. Therefore, the linearization performance of the algorithm is evaluated by the baseband MSE and the out-of-band adjacent channel power ratio, where the adjacent channel power ratio is the ratio of the average power of the adjacent bandwidth to the average power of the signal bandwidth.
[0045] In this scenario, a broadband OFDM signal with a bandwidth of 2 MHz is used as the input excitation signal, and the signal is input into the baseband signal acquisition and digital predistortion module. Figure 4 The process achieves the convergence of the DPD algorithm, and the MSE and ACPR performance results of the broadband signal scenario are shown in Table 2. It can be concluded from Table 2 that the joint optimization algorithm can also obtain an 8dB mean square error gain and a 5-6dB adjacent channel power ratio gain on the basis of the existing DPD.
[0046] Table 2 Performance indicators of various methods in multi-tone scenarios
[0047]
[0048]
[0049] In summary, the present invention uses a baseband-RF joint optimization algorithm to further optimize out-of-band leakage such as IMD3 on the basis of MSE reduction, and successfully improves system performance. By directly observing the out-of-band leakage in the spectrum diagram through the joint algorithm to measure the linearization performance, the index requirements of the RF end analog pre-distortion in the actual system can be met, such as Figure 5 The present invention can be applied to narrowband scenarios of baseband digital dual-tone signals, and can also be applied to broadband multi-tone signal scenarios.
[0050] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.
Claims
1. A digital predistortion method based on baseband-radio frequency joint optimization, characterized in that: The data set of transmit and receive signals of a real RF power amplifier is collected to train the power amplifier model to obtain the model coefficient vector. The data set of baseband excitation signals is collected and input into the digital pre-distortion model, and then input into the power amplifier model after up-conversion processing to obtain the corresponding RF output signal and its power spectrum diagram. The RF output signal is then down-converted and linearly filtered to obtain the baseband recovery signal, and then the feedback signal is obtained through the post-distortion model to calculate the error and MSE between the excitation and feedback. The mathematical expression of the determining factor of the nonlinear intermodulation distortion is calculated according to the model coefficient vector of the digital pre-distortion model and the model coefficient vector of the power amplifier model and the up-down conversion related parameters. Finally, the recursive least squares (RLS) improved algorithm is derived according to the joint optimization objective function based on the baseband-RF characteristics, and the digital pre-distortion model coefficient vector is iteratively updated according to the algorithm, so that the joint optimization objective function converges to achieve the linear optimization of the RF power amplifier, so that the digital index MSE and the RF index IMDs are reduced together. The power amplification model is: RF (n) = b1z RF (n)+b2z RF (n) 2 +b3z RF (n) 3 , where: b = [b1, b2, b3] is the model coefficient vector, y RF (n) is the output radio frequency signal; The digital pre-distortion model is: z(n)=a1x(n)+a2x(n) 2 +a3x(n) 3 , where: a = [a1, a2, a3] is the pre-distorter model coefficient, x(n) is the baseband input signal; the memoryless model of the post-distorter: Where: y(n) is the baseband output signal; The input and output model of the entire link end-to-end at the transmitter is: Where: x(n) is the baseband input signal, and the coefficients of the odd terms are: c3=a3b1cosω c +2a1a2b2cosω c 2 +a1 3 b3cosω c 3 , c5=2a2a3b2cosω c 2 +(3a1 2 a3+a1a2 3 )b3cosω c 3 , c7=(3a1a3 2 +3a2 2 a3)b3cosω c 3 ; The joint optimization objective function based on baseband-RF characteristics is: Obj=MSE+IMDs, where: IMDs=αc3+βc5+γc7, α, β, γ are weight coefficients respectively; Where N is the sample length of the data set, and since the predistorter and postdistorter provide the same input and output characteristics, y(n)-x(n) can be equivalent to the error 2. The digital predistortion method based on baseband-RF joint optimization according to claim 1 is characterized in that: include: Step 1: Based on the change of the baseband signal, evaluate whether the end-to-end system needs to add a new digital pre-distortion model, or whether the trained digital pre-distortion model is suitable for the current excitation signal. Make a judgment based on user needs and baseband signal configuration. The specific judgment criteria and methods are: observe the AM-AM characteristics of the PA transceiver, the NMSE performance of the baseband transceiver data set, and the out-of-band leakage of the RF signal frequency domain; Step 2: Initialize all parameters in the adaptive algorithm, including: forgetting factor λ, regularization parameter δ, inverse correlation matrix P(n), and IMDs expression control factors α, β, and γ for controlling the stability of the system, the convergence speed and accuracy of the algorithm; initialize the digital pre-distortion model coefficient vector a(0) so that the digital pre-distortion model provides the simplest linear change for the excitation signal in the initial state; the baseband signal is converted to the RF section after passing through the DPD initial model and output to the real RF power amplifier. The corresponding sampling frequency is configured according to the bandwidth of the baseband signal and the carrier frequency of the up and down conversion, and the RF signals sent and received at both ends of the PA are sampled to train the power amplifier model and obtain the model coefficient vector b; Step 3: Update the digital pre-distortion model according to a(n), input the baseband signal into the digital pre-distortion model, and then pass through the DAC, up-convert to the RF section, and then output to the real RF power amplifier to increase the signal transmission power and generate nonlinear distortion; Step 4: down-convert the PA output RF signal, obtain the baseband output signal from the ADC and transmit it to the post-distorter, which generates a feedback signal for indicator calculation and algorithm iteration; Step 5: According to the coefficient vector of the digital pre-distortion model and the power amplifier model and the digital up-conversion characteristics, the expressions of the odd-order coefficients of the end-to-end model are calculated, the mathematical expression of the IMDs control factor is derived, and the system error and the mean square error MSE are calculated according to the feedback signal and the excitation signal. The MSE and IMDs indicators are used to construct a jointly optimized MSE+IMDs objective function, and the objective function is gradually iterated and converged to the least squares solution through the derived improved recursive least squares adaptive algorithm, which specifically includes: Step 5.1, calculating the gain vector matrix k(n) according to the inverse correlation matrix, the forgetting factor, and the feedback signal to control the iterative update variation of the coefficient vector of the digital predistortion model; Step 5.2, calculate the system objective function, and update the expressions of system error e(n) and IMDs control factor; Step 5.3, iteratively updating the digital predistortion model coefficient vector a(n) according to the gain vector, forgetting factor, inverse correlation matrix, system error, IMDs and their control factors; Step 5.4, update the inverse correlation matrix; Step 5.5: Determine whether the joint optimization index converges. If not, repeat steps 3 to 5.
5. Step 6: Apply the converged digital pre-distortion model to the transceiver system, linearize the excitation signal, calculate the linearity index to evaluate the algorithm performance, and measure the RF signal spectrum of the RF power amplifier output to observe the out-of-band distortion.
3. A digital predistortion device based on baseband-RF joint optimization for implementing the method described in claim 1 or 2, characterized in that: include: The baseband signal acquires a digital predistorter module, a radio frequency front-end module, a post-distorter feedback signal generation module, and an objective function construction adaptive algorithm iteration module, wherein: the baseband signal acquires a digital predistorter module to construct a digital predistortion model according to the time domain data set of the excitation signal of the baseband transmitter and the DPD coefficient vector from the objective function construction adaptive algorithm iteration module, which is used to add predistortion characteristics to the excitation signal, and then outputs the processed excitation signal to the radio frequency front-end module and the objective function construction adaptive algorithm iteration module respectively according to the indirect learning structure; the radio frequency front-end module converts the baseband predistortion signal to the radio frequency segment, amplifies the signal power, and transmits it through the antenna at the same time The data set is sampled and down-converted and output to a post-distorter feedback signal generation module; the post-distorter feedback signal generation module generates a feedback signal according to the down-converted transmission signal and outputs it to an objective function construction adaptive algorithm iteration module; the objective function construction adaptive algorithm iteration module obtains the baseband signal of the digital pre-distorter module from the baseband signal and the feedback signal from the post-distorter feedback signal generation module to calculate the system error and the digital indicator MSE, and constructs an end-to-end input and output model of the overall link of the transmitter according to the vector of the digital pre-distortion model and the vector of the digital up-conversion and power amplification model, so as to jointly linearize the in-band and out-of-band distortions of the digital pre-distortion model.
4. The digital predistortion device according to claim 3, characterized in that: The RF front-end module includes: up-converter, down-converter, A / D converter and real RF power amplifier, wherein: the baseband signal is converted to the RF section by the up-converter and DAC and then outputted. RF (n) = v (cosω1n + cosω2n) to the real RF power amplifier, which generates system nonlinear distortion while performing power amplification. The real RF power amplifier outputs the RF signal y RF (n) The signal is transmitted through the antenna and output to the post-distortion feedback signal generation module, and then is processed by the ADC, down-converted back to the baseband, and processed by the post-distortion device to generate a feedback signal.
5. The digital predistortion device according to claim 3, characterized in that: The baseband signal acquisition digital predistorter module comprises: a baseband signal acquisition unit and a digital predistorter generation and update unit, wherein: the baseband signal acquisition unit is connected to the baseband physical layer module and receives the OFDM signal generated by it, the digital predistorter generation and update unit generates an initialized DPD model, and then updates the DPD model according to the DPD model coefficients iterated by the adaptive algorithm iteration module constructed according to each objective function, until the joint optimization target meets the required threshold, at which time the final DPD model is generated and applied to the transceiver system to obtain a baseband signal result containing predistortion.
6. The digital predistortion device according to claim 3, characterized in that: The post-distorter feedback signal generating module comprises: a down-conversion unit and a post-distorter unit, wherein: the down-conversion unit is connected to the RF front-end module and samples and transmits RF transmission signal information, and down-converts the RF signal to baseband; the post-distorter unit performs post-distortion adding processing according to the transmission signal information after down-conversion; the post-distorter is a pre-distortion model having the same input and output characteristics as the digital pre-distortion to obtain the feedback signal data set result.
7. The digital predistortion device according to claim 3, characterized in that: The objective function construction adaptive algorithm iteration module includes: a system error and digital indicator calculation unit, an analog indicator IMD3 mathematical expression calculation unit and an improved RLS adaptive algorithm iteration unit, wherein: the system error and digital indicator calculation unit is connected to the baseband signal acquisition digital predistorter module and the post-distorter feedback signal generation module and obtains the baseband predistortion signal and the feedback predistortion signal, the difference between the two is the system nonlinear error, and the digital indicator MSE is calculated based on this; the analog indicator IMD mathematical expression calculation unit performs IMD3 expression calculation according to the current DPD model coefficient, the up-conversion modulation formula and the PA model coefficient to obtain the equivalent formula of the analog indicator IMD; the improved RLS adaptive algorithm iteration unit jointly optimizes the objective function by jointly designing the digital / analog indicators, derives the model coefficient recursive formula of the improved adaptive algorithm, performs iterative update calculation of the DPD model coefficient, and finally obtains the coefficient least squares solution that makes the objective function converge to the specified threshold.
8. The digital predistortion device according to claim 3, characterized in that: The joint linearization of in-band and out-of-band distortions refers to: calculating the IMDs mathematical expression as an analog indicator, thereby constructing a joint optimization objective function based on baseband-RF characteristics based on the digital / analog indicator, and iterating the digital pre-distortion model coefficient vector a using a new improved RLS adaptive algorithm derived from the joint optimization objective function, and outputting it to the baseband signal to obtain the digital pre-distortion module to update the digital pre-distortion model in real time. When the joint indicator converges, the iterative update is stopped. At this time, the generated digital pre-distortion model can realize the joint linearization of the system's in-band and out-of-band distortions.
