A method for implementing a digital predistorter for 5G wireless networks

Abstract

The present invention discloses a method for implementing a digital predistorter for a 5G wireless network. The method first sets the MP model as a predistorter model, collects baseband signals, first baseband data and first output data, and then reconstructs the MP model in sections, optimizes and updates the section coefficients using a secondary synchronous perturbation random approximation algorithm, and calculates the correlation coefficient of the predistorter based on the optimized section coefficients combined with the first baseband data and the first output data. Finally, the correlation coefficient is multiplied by the baseband signal to obtain the second baseband data, and the second baseband data is input into a power amplifier to obtain the second output data. The present invention improves the linearization performance of the power amplifier, and can achieve the effect of maximizing the predistortion performance of the predistorter when 5G resources are limited.

Description

Technical Field

[0001] The present invention relates to the field of 5G wireless communication technology, and in particular to a method for implementing a digital predistorter for a 5G wireless network. Background Art

[0002] In wireless communication systems, the RF power amplifier in a 5G transmitter is one of the essential devices. Due to the inherent nonlinearity and memory effects of the RF terminal power amplifier, the transmitter efficiency is low and the quality of the transmitted signal deteriorates. In order to meet the high-efficiency transmission indicators and high-quality requirements of signal transmission, it is necessary to compensate for the nonlinearity and memory effects of the RF terminal power amplifier.

[0003] Digital pre-distortion (DPD) is one of the most mainstream linearization technologies with the best correction effect. Digital pre-distortion is the pre-distortion processing of digital signals at the baseband. Its pre-distortion processing simulates the inverse function model of the power amplifier. After the signal is pre-distorted and amplified by the power amplifier, the linear output of the signal is guaranteed.

[0004] The commonly used pre-distortion model in the industry is the memory polynomial. With the expansion of 5G signal bandwidth, the memory polynomial has a good compensation ability for strong nonlinearity and deep memory effect. However, the more segments there are, the higher the model fitting accuracy is, and more digital processing resources are required to realize the operation. Therefore, in view of the limited resources of 5G, it is necessary to achieve the best performance of digital pre-distortion with a smaller number of segments. Summary of the invention

[0005] In order to improve the linearization performance of a power amplifier with fewer resources, the present invention provides a method for implementing a digital predistorter for a 5G wireless network.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0007] A method for implementing a digital predistorter for a 5G wireless network includes setting an MP model as a predistorter model, and further comprising the following steps:

[0008] S1, capturing a baseband signal, first baseband data and first output data of a predistorter;

[0009] S2, reconstruct the MP model in sections, optimize the section coefficients and update them;

[0010] S3, calculating a correlation coefficient of a predistorter based on the first baseband data, the first output data and the optimized segment coefficient;

[0011] S4, multiplying the correlation coefficient by the baseband signal to obtain second baseband data, and inputting the second baseband data into a power amplifier to obtain second output data.

[0012] Preferably, a method for implementing a digital predistorter for a 5G wireless network, the segmented formula of the MP model is as follows:

[0013]

[0014] in, is the baseband signal, is the first output data, K is the number of segments, k is the segment cycle factor, M is the memory depth, m is the initial memory depth, is the correlation coefficient, β k is the segmentation coefficient.

[0015] Preferably, a method for implementing a digital predistorter for a 5G wireless network, wherein the segment coefficient is a quadratic function of a root mean square error and is optimized using a quadratic synchronous perturbation random approximation algorithm.

[0016] Preferably, a method for implementing a digital predistorter for a 5G wireless network, updating the segment coefficients comprises the following steps:

[0017] 101, capture data;

[0018] 102, generated and

[0019] 103. Calculate the loss function and L(B j );

[0020] 104, Update B using the quadratic synchronous perturbation random approximation algorithm j+1 vector;

[0021] 105, determine whether the maximum number of iterations has been reached,

[0022] If reached, the update is completed;

[0023] If not reached, steps 101-104 are repeated until the update is completed.

[0024] Preferably, a method for implementing a digital predistorter for a 5G wireless network generates and Using the following formula,

[0025]

[0026]

[0027] Bj =[β1,β2,…β k ];

[0028] In the formula, B j For the segmentation coefficient β k The segment vector of j is a constant that controls the convergence speed; Δ j is the random perturbation vector, Δ j =[1,-1,…,1], and Δ j Satisfies Bernoulli distribution.

[0029] Preferably, a method for implementing a digital predistorter for a 5G wireless network, updating B j+1 The vector uses the following formula,

[0030]

[0031] Preferably, in a method for implementing a digital predistorter for a 5G wireless network, in step S3, a method for calculating the correlation coefficient of the predistorter adopts a least squares method.

[0032] Compared with the prior art, the present invention has the following beneficial effects: in the pre-distorter model, the MP model is segmentedly reconstructed, the segmented coefficients in the MP model are optimized, and the optimized segmented coefficients are brought into the pre-distorter model, and the pre-distortion is completed to obtain the second output data, so that the linearization performance of the power amplifier is better, and the pre-distortion performance of the pre-distorter can be maximized when 5G resources are limited. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is the principle block diagram of DPD;

[0034] Figure 2 Update the flow chart for adaptively updating the segment coefficients;

[0035] Figure 3 It is a comparison chart of the uniformly distributed segment coefficient and the optimized segment coefficient NMSE;

[0036] Figure 4 The ACPR comparison chart of the uniformly distributed segment coefficient and the optimized segment coefficient based on 100MHz bandwidth; DETAILED DESCRIPTION

[0037] The present invention is further described in detail below in conjunction with test examples and specific implementation methods. However, this should not be understood as the scope of the above subject matter of the present invention being limited to the following embodiments, and all technologies realized based on the content of the present invention belong to the scope of the present invention.

[0038] Example 1

[0039] A method for implementing a digital predistorter for a 5G wireless network includes setting an MP model as a predistorter model. In this embodiment, the segmented formula of the MP model is as follows:

[0040]

[0041] is the baseband signal, is the first output data, K is the number of segments, k is the segment loop factor, and the loop range is [1:K]; M is the memory depth, m is the initial memory depth, and the loop range is [0:M]; is the correlation coefficient, β k is the segmentation coefficient.

[0042] Reference Figure 1 and Figure 2 After the setting is completed, the following steps are also included:

[0043] S1, capturing a baseband signal, first baseband data and first output data of a predistorter;

[0044] The baseband signal is Figure 1 In It is a complex baseband signal; the first baseband data is Figure 1 In The first output data is Figure 1 In It is the baseband signal after predistortion.

[0045] S2, reconstruct the MP model in sections, optimize the section coefficients and update them;

[0046] In this embodiment, the quadratic synchronous perturbation random approximation algorithm is used to optimize the segmentation coefficient β k , refer to Figure 2 , the optimization of the piecewise coefficients by the quadratic synchronous perturbation random approximation algorithm includes the following steps:

[0047] 101, capture data;

[0048] Capture the first output data of the predistorter

[0049] 102, generated and

[0050] Based on the captured first output data Get the segment coefficient β k , let the segment vector B j =[β1,β2,…,β k ], and Use the following formula:

[0051]

[0052]

[0053] In the formula, j is the number of iterations, b is j is a constant that controls the convergence speed; Δ j is the random perturbation vector, Δ j =[1,-1,…,1], and Δ j Satisfies Bernoulli distribution.

[0054] 103. Calculate the loss function and L(B j );

[0055] The loss function is calculated using the root mean square error (NMSE) formula of the power amplifier. The calculation formula is:

[0056]

[0057] in, Simulation data constructed for formula (1);

[0058] From formula (5), we can know that the segmentation coefficient β k It is a quadratic function of the root mean square error, that is, B j , and The loss functions are expressed as and L(B j ).

[0059] 104, Update B using the quadratic synchronous perturbation random approximation algorithm j+1 vector;

[0060] Specifically, update B j+1 The vector uses the following formula:

[0061]

[0062] 105, determine whether the maximum number of iterations has been reached, if so, complete the update; if not, repeat steps 101-104 until the update is completed.

[0063] S3, based on the first baseband data First output data And the optimized segmentation coefficient B j+1 , calculate the correlation coefficient of the predistorter

[0064] Based on the first output data The inverse function model of the power amplifier is obtained. This model has good compensation ability for strong nonlinearity and deep memory effect. Combined with the first baseband data And the optimized segmentation coefficient B j+1 , the least squares method (LS algorithm) is used to extract the correlation coefficient

[0065] S4, the correlation coefficient With baseband signal The second baseband data is multiplied to obtain second baseband data, and the second baseband data is input into a power amplifier (PA) to obtain second output data.

[0066] Example 2

[0067] After 100 iterations in this embodiment, i.e., j=100, the NMSE comparison of the uniformly distributed segment coefficient and the optimized segment coefficient is shown in the figure below: Figure 3 As shown in the figure, it can be seen that when the two algorithms converge respectively, the optimized segment coefficient is about 10dB smaller than the NMSE value of the uniform segment coefficient; the ACPR comparison of the uniformly distributed segment coefficient and the optimized segment coefficient in 100MHz bandwidth is shown in Figure 4 As shown in the figure, it can be seen that the ACPR value of the optimized segment coefficient is about 10dB smaller than that of the uniformly distributed segment coefficient.

[0068] From the test data table 1, it can be seen that under the same number of segmentation coefficients, the linearization effect of the optimized segmentation coefficients is better than that of the uniform segmentation coefficients.

[0069] Table 1

[0070] No pre-distortion Uniform distribution segment coefficient Optimize segmentation coefficient NMSE(dB) -24 -35 -46 ACPR(dBc) -28 / -29 -40 / -39 -50 / -51

[0071] In summary, the quadratic synchronous perturbation random approximation algorithm is used to adaptively optimize the segmentation coefficients, and the best performance of digital pre-distortion is achieved under a smaller number of segments. Compared with the uniform segmentation coefficients, the optimized segmentation coefficients make the linearization performance of the power amplifier better, while ensuring the feasibility of the engineering implementation of the algorithm, and realizing the consumption of a large amount of digital signal processing (DSP) resources in exchange for a small amount of storage resources, which is a better method for engineering implementation.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for implementing a digital predistorter for a 5G wireless network, comprising setting an MP model as a predistorter model, characterized in that: The following steps are included: S1, capturing a baseband signal, first baseband data and first output data of a predistorter; S2, reconstruct the MP model in sections, optimize the section coefficients and update them; The piecewise formula of the MP model is as follows, ； in, is the baseband signal, is the first output data, K is the number of segments, k is the segment cycle factor, M is the memory depth, m is the initial memory depth, is the correlation coefficient, β k is the segmentation coefficient; The quadratic synchronous perturbation random approximation algorithm is used to optimize the segmentation coefficient β k , the optimization of the piecewise coefficients by the quadratic synchronous perturbation random approximation algorithm includes the following steps: 101, capture data; capture the first output data of the predistorter ; 102, generated and ; Based on capturing the first output data of the predistorter , and obtain the segmentation coefficient β k , B j For the segmentation coefficient β k The segment vector of j =[β1,β2,…,β k ], and Using the following formula, ； In the formula, j is the number of iterations, b is j is a constant that controls the convergence speed; Δ j is the random perturbation vector, Δ j =[1,-1,…,1], and Δ j Satisfies Bernoulli distribution; 103. Calculate the loss function , and L(B j ); Segment coefficient β k is a quadratic function of the root mean square error, Bj, and The loss functions are expressed as , and L(B j ); The loss function is calculated using the root mean square error (NMSE) formula of the power amplifier. The calculation formula is: ； in, Simulation data constructed for the piecewise formulation of the MP model; 104, Update B using the quadratic synchronous perturbation random approximation algorithm j+1 vector; Update B j+1 The vector uses the following formula, ； 105, determine whether the maximum number of iterations has been reached, If reached, the update is completed; If not reached, repeat steps 101-104 until the update is completed; S3, calculating the correlation coefficient of the predistorter based on the first baseband data, the first output data and the optimized segment coefficient; S4, multiplying the correlation coefficient by the baseband signal to obtain second baseband data, and inputting the second baseband data into a power amplifier to obtain second output data.

2. The method for implementing a digital predistorter for a 5G wireless network according to claim 1, characterized in that: In step S3, the correlation coefficient of the predistorter is calculated by the least square method; An inverse function model of the power amplifier is obtained based on the first output data, and then combined with the first baseband data and the optimized segmentation coefficient, a least square method is used to extract the correlation coefficient.

Images