Nonlinear modeling method for wideband transmitters

Abstract

The application discloses a wide-band transmitter nonlinear modeling method, and relates to the technical field of transmitter design. The method comprises the following steps: calculating a first model parameter value according to a baseband signal of a receiving end and a baseband signal of a sending end; obtaining an MP modeling model according to the first model parameter value; performing pre-distortion processing by using the MP modeling model to obtain an AM-AM curve and an AM-PM curve; obtaining two threshold values according to the two curves; obtaining a second model parameter value and a third model parameter value according to the first model parameter value and the baseband signal of the sending end; bringing the second model parameter value into a first GMP model to obtain a first GMP modeling model; bringing the third model parameter value into a second GMP model to obtain a second GMP modeling model; and obtaining a wide-band transmitter modeling model according to the three modeling models and the two threshold values. The application can accurately represent the nonlinear characteristics of the transmitter in a wide frequency band while reducing the complexity in the parameter extraction process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal transmitter design, and particularly relates to a nonlinear modeling method for a wideband signal transmitter. BACKGROUND

[0002] In a wireless communication system, with the application of larger bandwidth and high peak-to-average ratio signals, higher requirements are put forward for the linear performance of signals. Nonlinear distortion of a wideband signal transmitter is mainly caused by a power amplifier. After an input signal passes through the power amplifier, amplitude distortion and phase distortion are generated. The amplitude distortion is manifested as compression of the signal, and the phase distortion is manifested as a change in the phase difference between the input and output signals with the amplitude. In addition, new frequency components are generated, which improves the signal-to-noise ratio of other frequency bands and the communication quality of adjacent channels.

[0003] In addition to amplitude distortion and phase distortion, the power amplifier also has a memory effect. The memory effect is manifested in that the current output signal is related not only to the current input signal but also to the past input signal. Especially for a wideband input signal, with the increase of the bandwidth, the memory effect caused by the power amplifier is more obvious. Therefore, the nonlinear modeling of the power amplifier must take the memory effect into account.

[0004] The memory polynomial (MP) model is a common nonlinear model with a memory effect. The nonlinear characteristics of the power amplifier are represented by deleting cross terms with inconsistent memory depths in the Volterra series, as shown in formula (1). The MP model has the advantages of simple implementation and high flexibility, and can compensate for most nonlinear distortion caused by the power amplifier.

[0005]

[0006] wherein y mp is the output of the MP model, x(n) is the input of the MP model, x(n-q) is a delay of x(n) by q units, k represents the kth order of |x(n)|, K is the nonlinear order, Q is the memory depth, akq is a model parameter, and || is a complex absolute value operation. The MP model has the advantages of high flexibility and simple implementation, and can model a simple power amplifier. However, the performance of the MP model decreases when the memory effect is strong.

[0007] The input signal with greater bandwidth, higher peak-to-average ratio and more complex power amplifier brings nonlinear distortion with stronger memory effect which is difficult to compensate with memory polynomial model. Therefore, a generalized memory polynomial (GMP) model is proposed to solve the predistortion model problem of wideband signal, which improves the linearization performance by adding additional cross terms with different memory depths on the basis of MP, as shown in formula (2).

[0008]

[0009] wherein K is a nonlinear order, Q is a memory depth, a kq , bklq, ck lq are model parameters, || is a complex absolute value operation, y gmp represents the GMP model output, -l represents a delay of l units, +l represents an advance of l units, Q b represents the memory depth of the second term, Q c represents the memory depth of the third term, L b is a lag depth, L c is an advance depth.

[0010] Although the GMP model can better model the complex nonlinear behavior of the memory effect of the wideband signal, with the addition of cross terms and the increase of memory depth, the complexity of the GMP model in the parameter extraction process increases sharply. SUMMARY

[0011] The purpose of the present application is to provide a wideband transmitter nonlinear modeling method which can accurately represent the nonlinear characteristics of the transmitter in a wide frequency band while reducing the complexity in the parameter extraction process.

[0012] To achieve the above purpose, the present application provides the following scheme:

[0013] A wideband transmitter nonlinear modeling method, comprising:

[0014] An offline experiment platform is constructed to obtain the baseband signal of the receiving end and the baseband signal of the sending end of the offline experiment platform;

[0015] A digital predistortion model is constructed;

[0016] The first model parameter value in the digital predistortion model is calculated according to the baseband signal of the receiving end and the baseband signal of the sending end; the first parameter value is the model parameter value of the MP model;

[0017] The first model parameter value is brought into the MP model to obtain an MP modeling model;

[0018] The MP modeling model is used for pre-distortion processing of the baseband signal of the sending end, so as to obtain a pre-distorted AM-AM curve and a pre-distorted AM-PM curve.

[0019] The first segment threshold and the second segment threshold in the digital pre-distortion model are obtained according to the pre-distorted AM-AM curve and the pre-distorted AM-PM curve.

[0020] The second model parameter value and the third model parameter value in the digital pre-distortion model are obtained according to the first model parameter value and the baseband signal of the sending end; the second model parameter value is a model parameter value of a first GMP model, and the third model parameter value is a model parameter value of a second GMP model.

[0021] The first GMP modeling model is obtained by inputting the second model parameter value into the first GMP model.

[0022] The second GMP modeling model is obtained by inputting the third model parameter value into the second GMP model.

[0023] The wideband transmitter modeling model is obtained according to the MP modeling model, the first segment threshold, the second segment threshold, the first GMP modeling model and the second GMP modeling model, and the wideband transmitter modeling model is used for nonlinear modeling of the wideband transmitter.

[0024] Optionally, the digital pre-distortion model is specifically as follows:

[0025] Wherein, y represents a digital pre-distortion model, F gmp1 (·) represents a first GMP model, F mp (·) represents an MP model, F gmp2 (·) represents a second GMP model, λ1 represents a first segment threshold, λ2 represents a second segment threshold, and x(n) represents a baseband signal of a sending end.

[0026] Optionally, the offline experiment platform comprises a PC end, a vector signal generator, a power amplifier and a vector signal analyzer.

[0027] The PC end is used for generating an original baseband signal of the sending end and transmitting the original baseband signal of the sending end to the vector signal generator.

[0028] The vector signal generator is used for performing up-conversion operation on the original baseband signal of the sending end to obtain a radio frequency signal, and transmitting the radio frequency signal to the power amplifier.

[0029] The power amplifier is used for power amplification operation on the up-converted signal, to obtain a power-amplified radio frequency signal, and to send the power-amplified radio frequency signal to the vector signal analyzer.

[0030] The vector signal analyzer is used for down-conversion of the power-amplified radio frequency signal to obtain the original baseband signal of the receiving end.

[0031] The PC end is further used for alignment operation on the original baseband signal of the receiving end and the original baseband signal of the sending end, to obtain the baseband signal of the receiving end and the baseband signal of the sending end.

[0032] Optionally, the first model parameter value in the digital pre-distortion model is calculated according to the baseband signal of the receiving end and the baseband signal of the sending end, and specifically includes:

[0033] The baseband signal of the receiving end and the baseband signal of the sending end are brought into the MP model to obtain an MP equation.

[0034] The first model parameter value in the digital pre-distortion model is obtained by solving the MP equation.

[0035] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0036] By setting the digital pre-distortion model as a segmented function including the MP model, the first GMP model and the second GMP model, different models can be used for modeling according to different baseband signals, which can accurately represent the nonlinear characteristics of the transmitter in a wide frequency band while reducing the complexity of the parameter extraction process. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 Flow chart of the wide frequency band transmitter nonlinear modeling method of the present application;

[0039] Figure 2 Structural schematic diagram of the offline experiment in the present application;

[0040] Figure 3 Schematic diagram of the digital pre-distortion model in the present application;

[0041] Figure 4This is a flowchart of the pre-distortion parameter extraction process in this invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] like Figure 1 As shown, this embodiment of the invention provides a nonlinear modeling method for a wideband transmitter. The general steps are as follows: acquiring the baseband signals of the transmitter and receiver of the wideband transmitter, i.e., the offline experimental platform; selecting and constructing a digital predistortion model based on the distortion characteristics; modeling the power amplifier based on the baseband signals of the transmitter and receiver after delay alignment, and establishing a digital predistortion model and extracting parameters based on the power amplifier model. Based on the piecewise function fitting method, the basic idea of ​​this method is to segment the nonlinear distortion curve according to the power of the input signal, and use different functions to fit each segment of the curve to achieve modeling of complex nonlinear characteristics, as shown in equation (3).

[0045] y represents the digital predistortion model, i.e., the baseband signal at the receiver, where x(n) is the complex form of the input I / Q signal, i.e., the baseband signal at the transmitter, || represents the absolute value operation of the complex number, i.e., the amplitude of the input signal, and F M (·) The nonlinear fitting function for the Mth interval, where M is the number of segments, λ M For the threshold of segmentation, commonly used F(·) models include polynomial, MP, and DDR models.

[0046] The specific steps of the nonlinear modeling method for broadband transmitters include:

[0047] An offline experimental platform is constructed, and the baseband signal from the receiving end and the baseband signal from the transmitting end of the offline experimental platform are obtained.

[0048] Construct a digital predistortion model.

[0049] The first model parameter value in the digital predistortion model is calculated based on the baseband signal of the receiving end and the baseband signal of the transmitting end; the first parameter value is the model parameter value of the MP model.

[0050] The MP model is obtained by substituting the first model parameter values ​​into the MP model.

[0051] The MP model is used to predistort the baseband signal of the transmitting end to obtain the predistorted AM-AM curve and the predistorted AM-PM curve.

[0052] The first segment threshold and the second segment threshold in the digital predistortion model are obtained based on the predistorted AM-AM curve and the predistorted AM-PM curve.

[0053] The second and third model parameter values ​​in the digital predistortion model are obtained based on the first model parameter values ​​and the baseband signal of the transmitting end; the second model parameter values ​​are the model parameter values ​​of the first GMP model, and the third model parameter values ​​are the model parameter values ​​of the second GMP model.

[0054] The first GMP model is obtained by substituting the second model parameter values ​​into the first GMP model.

[0055] The third model parameter values ​​are substituted into the second GMP model to obtain the second GMP modeling model.

[0056] A broadband transmitter modeling model is obtained based on the MP modeling model, the first segment threshold, the second segment threshold, the first GMP modeling model, and the second GMP modeling model. The broadband transmitter modeling model is used to perform nonlinear modeling on the broadband transmitter.

[0057] In practical applications, such as Figure 2 As shown, the offline experimental platform includes a PC ( Figure 2 The image shows a PC, a vector signal generator, and a power amplifier. Figure 2 The image shows a power amplifier and a vector signal analyzer.

[0058] The PC is used to generate the original baseband signal of the transmitting end and transmit the original baseband signal of the transmitting end to the vector signal generator.

[0059] The vector signal generator is used to upconvert the original baseband signal at the transmitting end to obtain a radio frequency signal, and then send the radio frequency signal to the power amplifier.

[0060] The power amplifier is used to amplify the up-converted signal to obtain a power-amplified radio frequency signal, and then sends the power-amplified radio frequency signal to the vector signal analyzer.

[0061] The vector signal analyzer is used to down-convert the amplified radio frequency signal to obtain the original baseband signal at the receiving end.

[0062] The PC terminal is also used to perform an alignment operation on the original baseband signal of the receiving end and the original baseband signal of the transmitting end to obtain the baseband signal of the receiving end and the baseband signal of the transmitting end.

[0063] In practical applications, the specific working process of the offline experimental platform is as follows:

[0064] Step 1: Generate the baseband signal, i.e. the original baseband signal of the transmitting end, using MATLAB on the PC. The baseband signal includes: symbol rate, sampling rate, signal modulation method, and roll-off coefficient of raised cosine filter, etc.

[0065] Step 2: Connect the PC to the vector signal generator via the LAN port. Set the carrier frequency and download the baseband signal generated by MATLAB to the vector signal generator. The vector signal generator upconverts the signal to an RF signal and sends it to the power amplifier for amplification. The power amplifier outputs the amplified RF signal.

[0066] Step 3: The vector signal analyzer acquires the amplified RF signal output by the power amplifier, down-converts it to baseband to obtain the original baseband signal at the receiving end, and then uploads the original baseband signal at the receiving end to MATLAB on the PC via the LAN port and saves it.

[0067] Step 4: In MATLAB, the original baseband signal from the transmitting end and the original baseband signal from the output end are processed by a delay alignment algorithm to obtain the aligned baseband signal, which is the baseband signal from the receiving end and the baseband signal from the transmitting end.

[0068] In practical applications, such as Figure 3 As shown, the construction idea of ​​the digital predistortion model is as follows:

[0069] Step 1: Set two thresholds λ1 and λ2 for the baseband signal at the transmitting end according to the power level, where λ1 < λ2, and divide the baseband signal at the receiving end into three parts according to the thresholds.

[0070] Step 2: According to the phase distortion characteristics of the power amplifier, when the power of the baseband signal at the transmitting end is small, i.e., |x(n)|≤λ1, the memory effect of the power amplifier will cause strong phase distortion. Therefore, the GMP model with high memory depth and complex cross terms is selected as the predistortion function of the first part, as shown in formula (2).

[0071] Step 3: When the power of the baseband signal at the transmitting end is moderate, i.e., λ1≤|x(n)|≤λ2, the power amplifier works approximately in the linear region. The signal is amplified approximately linearly, and the memory effect brought by the power amplifier is relatively small. Therefore, the MP model with lower complexity can be selected to compensate for the nonlinear characteristics of this part, as shown in formula (1).

[0072] Step 4: When the power of the baseband signal at the transmitting end is relatively large, i.e., λ2≤|x(n)|, the output signal of the power amplifier begins to be significantly compressed, resulting in severe nonlinear distortion, while the memory effect has a relatively small impact. Therefore, in this part, a high-order and low-memory-depth GMP model can be selected. The GMP model is shown in formula (2). The high order and low memory depth are only different from the values ​​of K and Q corresponding to the model in step 3.

[0073] Step 5: The first term of the GMP model is set to the same nonlinear order and memory depth as the MP model, and the parameter coefficients extracted from the MP model are shared. Therefore, the digital predistortion model is shown in equation (4): Where y represents the output of the digital predistortion model, F gmp1 (·) represents the first GMP model, F mp (·) represents the MP model, F gmp2 (·) represents the second GMP model, λ1 represents the first segment threshold, λ2 represents the second segment threshold, and x(n) represents the baseband signal at the transmitting end.

[0074] In practical applications, the calculation of the first model parameter value in the digital predistortion model based on the baseband signal of the receiving end and the baseband signal of the transmitting end specifically includes:

[0075] The MP equations are obtained by substituting the baseband signals from the receiving end and the transmitting end into the MP model.

[0076] Solving the MP equation yields the first model parameter values ​​in the digital predistortion model.

[0077] In practical applications, the baseband signals from the receiving end and the transmitting end are substituted into the MP model to obtain the MP equation. Solving the MP equation yields the first model parameter values ​​in the digital predistortion model, specifically including:

[0078] The baseband signals at both the receiving and transmitting ends are normalized. In the digital predistortion model, the core MP part of the model is solved first. The baseband signal y(n) at the receiving end is used as the input of the model, and the baseband signal x(n) at the transmitting end is used as the ideal output of the model. The first model parameter values ​​are obtained by solving the model parameters through least squares.

[0079] In practical applications, the MP modeling model is used to predistort the baseband signal at the transmitting end, resulting in predistorted AM-AM and AM-PM curves. Based on these predistorted AM-AM and AM-PM curves, the first segmentation threshold and the second segmentation threshold in the digital predistortion model are obtained, such as... Figure 4 As shown, it specifically includes:

[0080] The baseband signal y1(n) at the receiving end is obtained by predistorting x(n) using the obtained MP modeling model (y1(n) is obtained by inverse operation input x(n)). The AM-AM and AM-PM curves after predistortion are plotted. Based on the AM-PM curve, the baseband signal at the transmitting end is divided into four intervals when the phase difference is less than 0.5: (0.05, 0.15), (0.15, 0.25), (0.25, 0.35), and (0.35, 0.45). The variance of the phase difference in each interval is calculated. When the variance of the phase difference between the current interval and the next interval is almost constant, the midpoint of the current interval is selected as λ1, and the variance σ... 2 The calculation is shown in formula (5): Where μ represents the mean of x(n) and N represents the total number of sample points.

[0081] Step 3: The selection of λ2 is mainly calculated based on existing methods. Specifically, it is based on the relationship between the slope and the rate of change of the slope of the AM-AM curve after pre-distortion of the MP model and the threshold. The threshold is mainly determined near the slope or the rate of change of the slope of 1. The selection of λ2 can also be obtained by other mathematical methods.

[0082] In practical applications, the second and third model parameter values ​​in the digital predistortion model are obtained based on the first model parameter values ​​and the baseband signal from the transmitting end. Specifically, this includes:

[0083] Find the parameters of the two piecewise functions (the first GMP model and the second GMP model) formed by the GMP model. For simplicity, let U... gmp Let the operator matrix formed by the cross terms of the GMP model be the parameter of the MP to be obtained, and let the parameter of the MP to be obtained also be the parameter of the first term of GMP. Then we can obtain (6) and (7). Solve for c by the least squares method. gmp :

[0084] U mp ·c mp +U gmp ·c gmp =x(n) (6)

[0085] U gmp ·c gmp =x(n)-U mp ·cmp (7)

[0086] Among them, U mp and U gmp The output signal y(n) of the normalized power amplifier is defined by the values ​​of the signal within a specific interval, which are known, and c mp c represents the parameters of the MP model obtained in the first part. gmp These are the coefficients of the other two terms in the GMP model.

[0087] The parameter solution process described in the article can be obtained by mathematical methods such as least squares method, least mean square error algorithm and recursive least squares method.

[0088] This invention constructs an offline experimental platform to record baseband signals from both the transmitting and receiving ends. Based on the amplitude and phase distortion characteristics of the power amplifier, a piecewise function (GMP and MP) is selected, with GMP reusing MP parameters to reduce parameter extraction complexity. Threshold λ1 is calculated using variance based on the AM-PM curve after MP predistortion, and threshold λ2 is calculated based on the AM-AM curve. Finally, the model undergoes two piecewise parameter extractions: first, the parameters of the MP model are solved as a whole, and the error after MP predistortion is calculated; second, the power of the baseband signal at the receiving end is determined, and the parameters of the second part of the GMP model are solved based on the thresholds. The parameters are then copied into the predistorter, whose model is consistent with the piecewise function solution model. By solving the parameters in the digital predistortion model, a broadband transmitter model is obtained, achieving nonlinear modeling of the broadband transmitter.

[0089] In this embodiment of the invention, the baseband signal at the receiving end is used as the input of the model, the baseband signal at the transmitting end is used as the output, and a suitable piecewise function digital predistortion model is selected. After predistortion by the piecewise function digital predistortion model, nonlinear modeling of the broadband transmitter is realized.

[0090] This invention employs the idea of ​​piecewise function fitting, aiming to accurately represent the nonlinear characteristics of a transmitter over a wide frequency band while requiring less complexity in model parameter extraction. Based on the piecewise function method, it improves the linearity of the linear transmitter by using MP or GMP piecewise, while reducing the complexity of parameter solving and solving the problem of continuity near the piecewise function. Furthermore, the model can dynamically adjust the threshold, exhibiting stronger robustness and enabling accurate modeling of nonlinear distortion in wideband transmitters. By measuring the signal of the wideband transmitter operating across the entire frequency band, it uses piecewise functions to dynamically select memory polynomials and generalized memory polynomials to fit the nonlinear distortion characteristics together with the input signal, providing a model foundation for linearization techniques such as transmitter pre-distortion correction.

[0091] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0092] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for modeling nonlinearity of a wideband transmitter, the method comprising: The application relates to a method for constructing a wideband transmitter modeling model. The method comprises the following steps: constructing an offline experiment platform, acquiring a baseband signal of a receiving end of the offline experiment platform and a baseband signal of a sending end; wherein y denotes a digital pre-distortion model, denotes a first GMP model, denotes an MP model, denotes a second GMP model, denotes a first segmentation threshold, denotes a second segmentation threshold, denotes a baseband signal at the transmitting end; constructing a digital pre-distortion model; the digital pre-distortion model is specifically: calculating a first model parameter value in the digital pre-distortion model according to the baseband signal of the receiving end and the baseband signal of the sending end; the first model parameter value is a model parameter value of an MP model; bringing the first model parameter value into the MP model to obtain an MP modeling model; adopting the MP modeling model to perform pre-distortion processing on the baseband signal of the sending end to obtain an AM-AM curve after pre-distortion and an AM-PM curve after pre-distortion; According to the first model parameter value and the baseband signal of the sending end, second model parameter value and third model parameter value in the digital pre-distortion model are obtained; the second model parameter value is a model parameter value of a first GMP model, and the third model parameter value is a model parameter value of a second GMP model; specifically, let U gmp be an operator matrix constituted by GMP model cross terms, and let the model parameter of the MP model to be solved also be a parameter of the first term of the GMP model, so that the formula and are obtained, and c gmp is solved by a least square method. U mp and U gmp is constituted by the value of the output signal of the normalized power amplifier y (n) in a specific interval, and is known. c mp is a parameter of the MP model, c gmp is a coefficient of the other two terms of the GMP model. obtaining a first segmentation threshold and a second segmentation threshold in the digital pre-distortion model according to the AM-AM curve after pre-distortion and the AM-PM curve after pre-distortion; bringing the second model parameter value into the first GMP model to obtain a first GMP modeling model; bringing the third model parameter value into the second GMP model to obtain a second GMP modeling model; 2. The wideband transmitter non-linear modeling method of claim 1, wherein, obtaining a wideband transmitter modeling model according to the MP modeling model, the first segmentation threshold, the second segmentation threshold, the first GMP modeling model and the second GMP modeling model; the wideband transmitter modeling model is used for nonlinear modeling of the wideband transmitter. The offline experiment platform comprises a PC end, a vector signal generator, a power amplifier and a vector signal analyzer; the PC end is used for generating an original baseband signal of the sending end and transmitting the original baseband signal of the sending end to the vector signal generator; the vector signal generator is used for performing up-conversion operation on the original baseband signal of the sending end to obtain a radio frequency signal and transmitting the radio frequency signal to the power amplifier; the power amplifier is used for performing power amplification operation on the radio frequency signal to obtain a power-amplified radio frequency signal and transmitting the power-amplified radio frequency signal to the vector signal analyzer; the vector signal analyzer is used for performing down-conversion on the power-amplified radio frequency signal to obtain an original baseband signal of the receiving end; 3. The wideband transmitter non-linear modeling method of claim 1, wherein, the PC end is further used for performing alignment operation on the original baseband signal of the receiving end and the original baseband signal of the sending end to obtain a baseband signal of the receiving end and a baseband signal of the sending end. The method for calculating the first model parameter value in the digital pre-distortion model according to the baseband signal of the receiving end and the baseband signal of the sending end specifically comprises the following steps: bringing the baseband signal of the receiving end and the baseband signal of the sending end into the MP model to obtain an MP equation; solving the MP equation to obtain the first model parameter value in the digital pre-distortion model.

Images