Dual-frequency power amplifier digital predistortion device and method based on two-dimensional IMSA function
By using the amplitude selection affine function of the two-dimensional instantaneous sample index in the dual-band amplifier digital predistortion model, the model structure and calculation process are simplified, the problem of high complexity of the dual-band amplifier behavior model is solved, and efficient predistortion is achieved Effect and low power consumption.
Patent Information
- Application Number
- CN202210281368.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-22
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-03-22
AI Technical Summary
The existing dual-band amplifier behavior model has high structure and computational complexity, making it difficult to achieve excellent linearization performance in actual hardware.
The amplitude-selected affine (IMSA) function based on two-dimensional instantaneous sample index is used to simplify the digital predistortion model of the dual-frequency power amplifier through segmentation and approximation methods, so that the basis function structure in each small area is linear, reducing the computational complexity.
It significantly reduces the computational complexity and hardware resource requirements, achieves excellent predistortion effects, reduces power consumption, and simplifies the difficulty of digital signal processing.
Smart Images

Figure CN114826168B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of digital predistortion of radio frequency power amplifiers, and in particular to a dual-frequency power amplifier digital predistortion device and method based on a two-dimensional instantaneous sample indexed amplitude selective affine (IMSA) function. Background Art
[0002] With the rapid development of modern wireless communication technology, users have higher and higher requirements for data transmission rate. Carrier aggregation technology can effectively combine scattered spectrum into a complete spectrum, and has become a common technology used to provide high data rate and spectrum utilization in modern digital communication systems. The non-continuous aggregation of multiple basic carriers between different frequency bands will produce multi-band signals. Under the concurrent excitation of multi-band signals, the output signal of the power amplifier (PA) often shows more complex distortion characteristics than the single-band case. At this time, the linearization performance of the traditional single-band digital pre-distortion (DPD) model will be significantly reduced.
[0003] Early research was conducted around dual-band PAs, and Bassam et al. pioneered the classic 2-D-DPD model, which performed two DPD processes by capturing the distortion information in two frequency bands separately. This model pointed the way for subsequent dual-band DPD research, but its structure was very complex. The two-dimensional modified MP (2-D-MMP) model successfully achieved similar complexity to single-band MP by using envelope coupling factors, but these factors needed to be searched through a binary search algorithm. The two-dimensional simplified MP (2-D-SMP) model eliminates the envelope coupling factors in the 2-D-MMP model and has lower complexity, but loses some modeling accuracy above the fifth order.
[0004] The above models can significantly compensate for the distortion in each frequency band and have excellent performance. However, their structures and calculation processes are mostly very complicated, which is not conducive to the implementation of actual hardware. Recently, a new method based on the amplitude-selective affine function (MSA) has been proposed. This method partitions and selects according to the amplitude of the signal sequence, so that the model structure in each partition becomes a linear structure, which greatly reduces the calculation complexity. This method has realized the DPD of single-band power amplifiers and shows excellent linearization ability. Summary of the invention
[0005] The present invention aims to reduce the complexity of the structure and calculation of the dual-band power amplifier behavior model in the above-mentioned prior art. The first object of the present invention is to provide a dual-band power amplifier digital predistortion device based on a two-dimensional instantaneous sample index amplitude selective affine (IMSA) function. At the same time, the second object of the present invention is to provide a dual-band power amplifier digital predistortion method based on a two-dimensional IMSA function based on the device.
[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:
[0007] The present invention provides a dual-frequency power amplifier digital predistortion device based on a two-dimensional IMSA function, comprising the following modules:
[0008] A dual-band power amplifier digital predistorter based on a two-dimensional instantaneous sample index amplitude selection affine (IMSA) function: The predistorter includes a dual-band power amplifier digital predistortion model based on a two-dimensional IMSA function, performs predistortion processing on a first-band input digital baseband signal x1(n) and a second-band input digital baseband signal x2(n), and generates a first-band digital predistortion signal x pre1 (n) and the second frequency band digital predistortion signal x pre2 (n);
[0009] The first digital-to-analog converter: digitally predistorts the first frequency band signal x pre1 (n) performing digital-to-analog conversion;
[0010] The second DAC: digitally predistorts the second frequency band signal x pre2 (n) performing digital-to-analog conversion;
[0011] The first modulator modulates the output signal of the first digital-to-analog converter to generate a radio frequency signal in a first frequency band;
[0012] The second modulator modulates the output signal of the second digital-to-analog converter to generate a radio frequency signal in a second frequency band;
[0013] Power combiner: combines the first frequency band RF signal and the second frequency band RF signal into one signal;
[0014] Power amplifier: amplifies and outputs the output signal of the power combiner;
[0015] Attenuation coupler: attenuates and couples the output signal of the power amplifier;
[0016] Power divider: divides the output signal of the attenuating coupler into two equal signals;
[0017] A first demodulator is used to demodulate a first frequency band of an output signal of the power divider;
[0018] A second demodulator is used to demodulate the second frequency band of the output signal of the power divider;
[0019] A first low-pass filter: performing a low-pass filter on the output signal of the first demodulator;
[0020] Second low-pass filter: performing low-pass filtering on the output signal of the second demodulator;
[0021] First analog-to-digital converter: performs analog-to-digital conversion on the output signal of the first low-pass filter to generate a first frequency band output digital baseband signal y1(n);
[0022] Second analog-to-digital converter: performs analog-to-digital conversion on the output signal of the second low-pass filter to generate a second frequency band output digital baseband signal y2(n);
[0023] Training of a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function: training of a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function is performed on the first-band input digital baseband signal x1(n), the second-band input digital baseband signal x2(n), the first-band output digital baseband signal y1(n), and the second-band output digital baseband signal y2(n) to generate a first-band predistortion coefficient and a second-band predistortion coefficient, and the first-band predistortion coefficient and the second-band predistortion coefficient are sent to a dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function.
[0024] Furthermore, the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function can be derived from the 2-D-DPD model by segmentation and approximation. When segmentation and approximation are performed according to the amplitudes of the two input signals of the model, its expression is shown in formula (1):
[0025]
[0026] Among them, the subscripts i=1 and 2 represent the first frequency band and the second frequency band respectively. is the basis function of the model in each small area after segmentation. M is the memory depth, L1 and L2 represent the number of partitions in the |y1(n)| and |y2(n)| dimensions respectively. and They represent the partition thresholds on the corresponding dimensions and can be preset. and They represent the three linear coefficients in each small area after segmentation when the memory depth is m. y1(nm) represents the digital baseband signal input to the first frequency band model at time nm, and y2(nm) represents the digital baseband signal input to the second frequency band model at time nm.
[0027] Furthermore, when the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is divided and approximated according to the square of the amplitude of the two input signals of the model, its expression is as shown in formula (2):
[0028]
[0029] Further, when the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is divided and approximated according to the amplitudes of the two input signals of the model, the specific derivation process is as follows:
[0030] 1. Establish the classic 2-D-DPD model expression in the concurrent dual-band scenario:
[0031]
[0032] Where M is the memory depth, K is the nonlinear order, is the model coefficient.
[0033] 2. Rewrite the 2-D-DPD model. The expression of the memoryless (M=1) 2-D-DPD model is:
[0034]
[0035] Where T (i) (|y1(n)|,|y2(n)|) are called the basis functions of the model in two frequency bands (i=1,2).
[0036] 3. Ignore the superscript i of the basis functions in the two frequency bands, and |y1(n)| and |y2(n)| are abbreviated as |y1| and |y2| respectively. According to the amplitude of the two input signals in the basis function, the two-dimensional plane is divided into several plane grids, and the basis function can be expressed as:
[0037]
[0038] After segmentation, the basis function on each small area of the plane can be obtained
[0039] 4. Performing a first-order Taylor series expansion of a binary function yields:
[0040]
[0041] in and is the linear term coefficient in each small area. Symbol O n represents the Peano-type remainder.
[0042] 5. When the entire two-dimensional plane is divided finely and densely enough, It can be approximated using only the linear terms in the first-order Taylor series expansion, O n can be ignored. At this time:
[0043]
[0044] 6. Formula (7) is similar in form to the result after using the MSA function, except that the dimension becomes two-dimensional. The model after considering the memory effect is recorded as the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function (2-D-MSA), and the specific expression is as follows:
[0045]
[0046] 7. Set the overall structure of the model to focus more on the impact of instantaneous samples and use instantaneous samples y in the constraints i (n) replaces the constraint y in formula (8) i (nm), the model is a dual-frequency power amplifier digital predistortion model (2-D-IMSA) based on the two-dimensional IMSA function when the amplitudes of the two input signals of the model are divided and approximated. The specific expression is as follows:
[0047]
[0048] Furthermore, when the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is divided and approximated according to the square of the amplitude of the two input signals of the model, the specific derivation process is similar to the above process, except that the basis function T in formula (4) is replaced when rewriting the 2-D-DPD model. (i) (|y1(n)|,|y2(n)|) are changed to the new basis function T (i) (|y1(n)| 2 ,|y2(n)| 2 ), and the amplitudes of the two input signals in the constraint conditions and Taylor expansion are replaced by the squares of the two signal amplitudes when the new basis functions are subsequently segmented and approximated. This model is a dual-frequency power amplifier digital predistortion model (2-D-IMSA) based on the two-dimensional IMSA function when segmenting and approximating the squares of the two input signal amplitudes of the model. The specific expression is as follows:
[0049]
[0050] Based on the above device, the dual-frequency power amplifier digital predistortion method based on the two-dimensional IMSA function described in the present invention comprises the following steps:
[0051] 1. Send the first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) to a dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function. The dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function includes a dual-frequency power amplifier predistortion model based on a two-dimensional IMSA function. The first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) are predistorted to generate a first frequency band digital predistortion signal x pre1 (n) and the second frequency band digital predistortion signal x pre2 (n);
[0052] 2. Digital predistortion signal x in the first frequency band pre1(n) performing digital-to-analog conversion and sending the converted signal to the first modulator to generate a radio frequency signal in the first frequency band; performing digital predistortion on the second frequency band signal x pre2 (n) performing digital-to-analog conversion and sending the result to a second modulator to generate a radio frequency signal in a second frequency band;
[0053] 3. Combine the first frequency band radio frequency signal and the second frequency band radio frequency signal into one signal and send it to the power amplifier for amplification and output;
[0054] 4. The output signal of the power amplifier is attenuated and coupled, and then passes through the power divider to be divided into two equal signals;
[0055] 5. The first output signal of the power divider is demodulated, filtered, and converted digitally to analog by a first demodulator, a first low-pass filter, and a first analog-to-digital converter in sequence to generate a first-band output digital baseband signal y1(n); the second output signal of the power divider is demodulated, filtered, and converted digitally to analog by a second demodulator, a second low-pass filter, and a second analog-to-digital converter in sequence to generate a second-band output digital baseband signal y2(n);
[0056] 6. Perform dual-frequency power amplifier digital pre-distortion model training based on the two-dimensional IMSA function on the first frequency band input digital baseband signal x1(n), the second frequency band input digital baseband signal x2(n), the first frequency band output digital baseband signal y1(n), and the second frequency band output digital baseband signal y2(n) to determine the first frequency band pre-distortion parameters and the second frequency band pre-distortion parameters.
[0057] Further, in step 6, the dual-frequency power amplifier digital predistortion model training based on the two-dimensional IMSA function includes the following steps:
[0058] 6.1. Obtain a first frequency band input digital baseband signal x1(n) and a first frequency band output digital baseband signal y1(n) and perform normalization and alignment processing; obtain a second frequency band input digital baseband signal x2(n) and a second frequency band output digital baseband signal y2(n) and perform normalization and alignment processing;
[0059] 6.2. When the model is segmented and approximated according to the amplitudes of the two input signals, formula (1) is used as the digital pre-distortion model; when the model is segmented and approximated according to the squares of the amplitudes of the two input signals, formula (2) is used as the digital pre-distortion model;
[0060] 6.3. Determine the memory depth M, the number of partitions L1 and L2 of the basis function in two dimensions, and the pre-set partition threshold and
[0061] 6.4. Based on the memory depth M, the number of partitions of the basis function in two dimensions L1 and L2 and the pre-set partition threshold and The first frequency band output digital baseband signal y1(n) and the second frequency band output digital baseband signal y2(n) are used as model input, and the first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) are used as model output to perform predistortion parameter training, and extract and determine the first frequency band predistortion parameters and the second frequency band predistortion coefficients.
[0062] Furthermore, in step 6.4, the specific pre-distortion coefficient extraction process includes the following steps:
[0063] 6.4.1. According to the constraints of the input y1(n) and y2(n) of the digital pre-distortion model, find the model input samples y1′(n) and y2′(n) and the corresponding model output samples x1′(n) and x2′(n) belonging to the same small plane area.
[0064] 6.4.2. The relationship between the input and output samples of the model in each small plane area can be written as:
[0065]
[0066] Where Y′ is the input matrix of the model, X′ is the corresponding model output vector, W′ is the model coefficient in each small area, N′ is the number of samples in each small area, and Q′=3M is the number of W′.
[0067] 6.4.3. Using the least squares (LS) algorithm, the corresponding solution can be expressed as
[0068]
[0069] in represents the extracted predistortion coefficient, (·) H represents the conjugate transpose of a complex matrix, (·) -1 Represents the inverse of the complex matrix. A dual-frequency power amplifier digital predistortion device and method based on a two-dimensional IMSA function of the present invention has the following advantages:
[0070] (1) The novel method of the present invention uses a two-dimensional IMSA function to replace the complex basis functions in the classical dual-band model through segmentation and approximation, so that the basis functions of the new model are simple linear structures in each small planar area.
[0071] (2) The 2-D-IMSA model described in the present invention does not require any high-order term operations during calculation, but only requires linear combinations between two dimensions. Under the premise of achieving excellent pre-distortion effects, the computational complexity is greatly reduced, and the hardware addition and multiplication resources and power consumption required by the model are significantly reduced.
[0072] (3) When the 2-D-IMSA model described in the present invention is divided and approximated according to the squares of the amplitudes of the two input signals of the model, a pipeline coordinate rotation digital calculation (CORDIC) module for square root calculation is not required, thereby reducing the difficulty and complexity of digital signal processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0073] Figure 1 It is a structural schematic diagram of a digital predistortion device in a specific implementation manner of the present invention;
[0074] Figure 2 is a schematic diagram of the segmentation process of the present invention;
[0075] Figure 3 The power spectrum of the power amplifier output signal before and after predistortion of two baseband signals with a bandwidth of 40 MHz, a peak-to-average ratio (PAPR) of about 7.0 dB, and a sampling rate of 245.76 MSPS using the 2-D-IMSA model in the method of the present invention when segmenting and approximating according to the amplitude of the two input signals of the model. (a) represents the first frequency band, and (b) represents the second frequency band. DETAILED DESCRIPTION
[0076] In order to better understand the purpose and advantages of the present invention, the technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.
[0077] This specific embodiment discloses a dual-frequency power amplifier digital predistortion device based on a two-dimensional IMSA function, such as Figure 1 As shown, it includes the following modules:
[0078] A dual-band power amplifier digital predistorter based on a two-dimensional instantaneous sample index amplitude selection affine (IMSA) function: The predistorter includes a dual-band power amplifier digital predistortion model based on a two-dimensional IMSA function, performs predistortion processing on a first-band input digital baseband signal x1(n) and a second-band input digital baseband signal x2(n), and generates a first-band digital predistortion signal x pre1 (n) and the second frequency band digital predistortion signal x pre2 (n);
[0079] The first digital-to-analog converter: digitally predistorts the first frequency band signal x pre1 (n) performing digital-to-analog conversion;
[0080] The second DAC: digitally predistorts the second frequency band signal x pre2 (n) performing digital-to-analog conversion;
[0081] The first modulator modulates the output signal of the first digital-to-analog converter to generate a radio frequency signal in a first frequency band;
[0082] The second modulator modulates the output signal of the second digital-to-analog converter to generate a radio frequency signal in a second frequency band;
[0083] Power combiner: combines the first frequency band RF signal and the second frequency band RF signal into one signal;
[0084] Power amplifier: amplifies and outputs the output signal of the power combiner;
[0085] Attenuation coupler: attenuates and couples the output signal of the power amplifier;
[0086] Power divider: divides the output signal of the attenuating coupler into two equal signals;
[0087] A first demodulator: demodulates a first frequency band of an output signal of the power divider;
[0088] A second demodulator is used to demodulate the second frequency band of the output signal of the power divider;
[0089] A first low-pass filter: performing a low-pass filter on the output signal of the first demodulator;
[0090] Second low-pass filter: performing low-pass filtering on the output signal of the second demodulator;
[0091] First analog-to-digital converter: performs analog-to-digital conversion on the output signal of the first low-pass filter to generate a first frequency band output digital baseband signal y1(n);
[0092] Second analog-to-digital converter: performs analog-to-digital conversion on the output signal of the second low-pass filter to generate a second frequency band output digital baseband signal y2(n);
[0093] Training of a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function: training of a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function is performed on the first-band input digital baseband signal x1(n), the second-band input digital baseband signal x2(n), the first-band output digital baseband signal y1(n), and the second-band output digital baseband signal y2(n) to generate a first-band predistortion coefficient and a second-band predistortion coefficient, and the first-band predistortion coefficient and the second-band predistortion coefficient are sent to a dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function.
[0094] Furthermore, the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function can be derived from the 2-D-DPD model by a segmentation and approximation method. When segmentation and approximation are performed according to the amplitudes of the two input signals of the model, its expression is shown in formula (1):
[0095]
[0096] Among them, the subscripts i=1 and 2 represent the first frequency band and the second frequency band respectively. is the basis function of the model in each small area after segmentation. M is the memory depth, L1 and L2 represent the number of partitions in the |y1(n)| and |y2(n)| dimensions respectively. and They represent the partition thresholds on the corresponding dimensions and can be preset. and They represent the three linear coefficients in each small area after segmentation when the memory depth is m. y1(nm) represents the digital baseband signal input to the first frequency band model at time nm, and y2(nm) represents the digital baseband signal input to the second frequency band model at time nm.
[0097] Furthermore, when the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is divided and approximated according to the square of the amplitude of the two input signals of the model, its expression is as shown in formula (2):
[0098]
[0099] Further, when the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is divided and approximated according to the amplitudes of the two input signals of the model, the specific derivation process is as follows:
[0100] 1. Establish the classic 2-D-DPD model expression in the concurrent dual-band scenario:
[0101]
[0102] Where M is the memory depth, K is the nonlinear order, is the model coefficient.
[0103] 2. Rewrite the 2-D-DPD model. The expression of the memoryless (M=1) 2-D-DPD model is:
[0104]
[0105] Where T (i) (|y1(n)|,|y2(n)|) are called the basis functions of the model in two frequency bands (i=1,2).
[0106] 3. Ignore the superscript i of the basis functions in the two frequency bands, and abbreviate |y1(n)| and |y2(n)| as |y1| and |y2| respectively. According to the amplitude of the two input signals in the basis function, the two-dimensional plane is divided into several plane grids. The division process is as follows: Figure 2 As shown. At this time, the basis function can be expressed as:
[0107]
[0108] After segmentation, the basis function on each small area of the plane can be obtained
[0109] 4. Performing a first-order Taylor series expansion of a binary function yields:
[0110]
[0111] in and is the linear term coefficient in each small area. Symbol O n represents the Peano-type remainder.
[0112] 5. When the entire two-dimensional plane is divided finely and densely enough, It can be approximated using only the linear terms in the first-order Taylor series expansion, O n can be ignored. At this time:
[0113]
[0114] 6. Formula (7) is similar in form to the result after using the MSA function, except that the dimension becomes two-dimensional. The model after considering the memory effect is recorded as the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function (2-D-MSA), and the specific expression is as follows:
[0115]
[0116] 7. Set the overall structure of the model to focus more on the impact of instantaneous samples and use instantaneous samples y in the constraints i (n) replaces the constraint y in formula (8) i (nm), the model is a dual-frequency power amplifier digital predistortion model (2-D-IMSA) based on the two-dimensional IMSA function when the amplitude of the two input signals of the model is divided and approximated. The expression is as follows:
[0117]
[0118] Furthermore, when the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is divided and approximated according to the square of the amplitude of the two input signals of the model, the specific derivation process is similar to the above process, except that the basis function T in formula (4) is replaced when rewriting the 2-D-DPD model. (i) (|y1(n)|,|y2(n)|) are changed to the new basis function T (i) (|y1(n)| 2 ,|y2(n)| 2), and the amplitudes of the two input signals in the constraint conditions and Taylor expansion are replaced by the squares of the two signal amplitudes when the new basis functions are subsequently segmented and approximated. This model is a dual-frequency power amplifier digital predistortion model (2-D-IMSA) based on the two-dimensional IMSA function when segmenting and approximating the squares of the two input signal amplitudes of the model. The specific expression is as follows:
[0119]
[0120] The specific implementation also discloses a dual-frequency power amplifier digital predistortion method based on a two-dimensional IMSA function, comprising the following steps:
[0121] 1. Send the first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) to a dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function. The dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function includes a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function. The first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) are predistorted to generate a first frequency band digital predistortion signal x pre1 (n) and the second frequency band digital predistortion signal x pre2 (n);
[0122] 2. Digital predistortion signal x in the first frequency band pre1 (n) performing digital-to-analog conversion and sending the converted signal to the first modulator to generate a radio frequency signal in the first frequency band; performing digital predistortion on the second frequency band signal x pre2 (n) performing digital-to-analog conversion and sending the result to a second modulator to generate a radio frequency signal in a second frequency band;
[0123] 3. Combine the first frequency band radio frequency signal and the second frequency band radio frequency signal into one signal and send it to the power amplifier for amplification and output;
[0124] 4. The output signal of the power amplifier is attenuated and coupled, and then passes through the power divider to be divided into two equal signals;
[0125] 5. The first output signal of the power divider is demodulated, filtered, and converted digitally to analog by a first demodulator, a first low-pass filter, and a first analog-to-digital converter in sequence to generate a first-band output digital baseband signal y1(n); the second output signal of the power divider is demodulated, filtered, and converted digitally to analog by a second demodulator, a second low-pass filter, and a second analog-to-digital converter in sequence to generate a second-band output digital baseband signal y2(n);
[0126] 6. Perform dual-frequency power amplifier digital pre-distortion model training based on the two-dimensional IMSA function on the first frequency band input digital baseband signal x1(n), the second frequency band input digital baseband signal x2(n), the first frequency band output digital baseband signal y1(n), and the second frequency band output digital baseband signal y2(n) to determine the first frequency band pre-distortion parameters and the second frequency band pre-distortion parameters.
[0127] Further, the dual-frequency power amplifier digital predistortion model training based on the two-dimensional IMSA function in step 6 includes the following steps:
[0128] 6.1. Obtain a first frequency band input digital baseband signal x1(n) and a first frequency band output digital baseband signal y1(n) and perform normalization and alignment processing; obtain a second frequency band input digital baseband signal x2(n) and a second frequency band output digital baseband signal y2(n) and perform normalization and alignment processing;
[0129] 6.2. When the model is segmented and approximated according to the amplitudes of the two input signals, formula (1) is used as the digital pre-distortion model; when the model is segmented and approximated according to the squares of the amplitudes of the two input signals, formula (2) is used as the digital pre-distortion model;
[0130] 6.3. Determine the memory depth M, the number of partitions L1 and L2 of the basis function in two dimensions, and the pre-set partition threshold and
[0131] 6.4. Based on the memory depth M, the number of partitions of the basis function in two dimensions L1 and L2 and the pre-set partition threshold and The first frequency band output digital baseband signal y1(n) and the second frequency band output digital baseband signal y2(n) are used as model input, and the first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) are used as model output to perform predistortion parameter training, and extract and determine the first frequency band predistortion parameters and the second frequency band predistortion coefficients.
[0132] Further, the specific pre-distortion coefficient extraction process in step 6.4 includes the following steps:
[0133] 6.4.1. According to the constraints of the input y1(n) and y2(n) of the digital pre-distortion model, find the model input samples y′1(n) and y′2(n) and the corresponding model output samples x′1(n) and x′2(n) belonging to the same small plane area.
[0134] 6.4.2. The relationship between the input and output samples of the model in each small plane area can be written as:
[0135]
[0136] Where Y′ is the input matrix of the model, X′ is the corresponding model output vector, W′ is the model coefficient in each small area, N′ is the number of samples in each small area, and Q′=3M is the number of W′.
[0137] 6.4.3. Using the least squares (LS) algorithm, the corresponding solution can be expressed as
[0138]
[0139] in represents the extracted predistortion coefficient, (·) H represents the conjugate transpose of a complex matrix, (·) -1 Represents the inverse of a complex matrix
[0140] The two baseband signals are both I / Q complex signals. The test uses two baseband signals with a bandwidth of 40MHz, a peak-to-average ratio (PAPR) of about 7.0dB, a sampling rate of 245.76MSPS, and a modulation frequency of 3.35GHz and 3.55GHz. The experiment tested the dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function when segmenting and approximating the amplitude of the two input signals of the model. The predistortion model parameters are determined as follows: memory depth M = 3, the number of partitions in the two dimensions L1 = L2 = 6, and the thresholds in the two dimensions are equally divided in the interval [0,1].
[0141] Table 1 shows the normalized mean square error (NMSE) during modeling and the adjacent channel leakage ratio (ACLR) before and after predistortion. Figure 3 The power spectrum of the power amplifier output in the two frequency bands before and after pre-distortion is plotted. It can be found that after adopting the dual-frequency power amplifier digital pre-distortion technology proposed in the present invention, the ACPR of the two sidebands of the output signal is reduced by 16 to 18 dBc, indicating that the nonlinearity of the power amplifier is well improved. From the above, it can be seen that the present invention simplifies the complex structure of the basis function of the 2-D-DPD model, and each small area after segmentation is a simple linear structure; at the same time, the 2-D-IMSA model in the present invention does not require any high-order term operations during calculation, but only requires linear combinations between two dimensions. Under the premise of achieving excellent pre-distortion effects, the calculation complexity is greatly reduced, and the hardware addition and multiplication resources and power consumption required for the model can be significantly reduced.
[0142] Table 1. NMSE and ACPR for two frequency bands
[0143]
Claims
1. A dual-frequency power amplifier digital predistortion device based on a two-dimensional IMSA function, characterized in that: include: A dual-band power amplifier digital predistortion model based on a two-dimensional IMSA function: Predistortion is performed on the first-band input digital baseband signal x1(n) and the second-band input digital baseband signal x2(n) to generate a first-band digital predistortion signal x pre1 (n) and the second frequency band digital predistortion signal x pre2 (n); The first digital-to-analog converter: digitally predistorts the first frequency band signal x pre1 (n) performing digital-to-analog conversion; The second DAC: digitally predistorts the second frequency band signal x pre2 (n) performing digital-to-analog conversion; The first modulator modulates the output signal of the first digital-to-analog converter to generate a radio frequency signal in a first frequency band; The second modulator modulates the output signal of the second digital-to-analog converter to generate a radio frequency signal in a second frequency band; Power combiner: combines the first frequency band RF signal and the second frequency band RF signal into one signal; Power amplifier: amplifies and outputs the output signal of the power combiner; Attenuation coupler: attenuates and couples the output signal of the power amplifier; Power divider: divides the output signal of the attenuating coupler into two equal signals; A first demodulator: demodulates a first frequency band of an output signal of the power divider; A second demodulator is used to demodulate the second frequency band of the output signal of the power divider; A first low-pass filter: performing a low-pass filter on the output signal of the first demodulator; Second low-pass filter: performing low-pass filtering on the output signal of the second demodulator; First analog-to-digital converter: performs analog-to-digital conversion on the output signal of the first low-pass filter to generate a first frequency band output digital baseband signal y1(n); Second analog-to-digital converter: performs analog-to-digital conversion on the output signal of the second low-pass filter to generate a second frequency band output digital baseband signal y2(n); Training of a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function: training of a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function is performed on the first-band input digital baseband signal x1(n), the second-band input digital baseband signal x2(n), the first-band output digital baseband signal y1(n), and the second-band output digital baseband signal y2(n) to generate a first-band predistortion coefficient and a second-band predistortion coefficient, and the first-band predistortion coefficient and the second-band predistortion coefficient are sent to a dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function.
2. A dual-frequency power amplifier digital predistortion device based on a two-dimensional IMSA function according to claim 1, characterized in that: The dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is derived from the 2-D-DPD model by segmentation and approximation. When segmentation and approximation are performed according to the amplitudes of the two input signals of the model, its expression is as shown in formula (1): Among them, the subscripts i=1 and 2 represent the first frequency band and the second frequency band respectively. is the basis function of the model in each small area after segmentation; M is the memory depth, L1 and L2 represent the number of partitions in the |y1(n)| and |y2(n)| dimensions respectively. and They represent the partition thresholds on the corresponding dimensions and can be preset; and They respectively represent the three linear coefficients in each small area after segmentation when the memory depth is m; y1(nm) represents the digital baseband signal input to the first frequency band model at moment nm, and y2(nm) represents the digital baseband signal input to the second frequency band model at moment nm.
3. The dual-frequency power amplifier digital predistortion device based on a two-dimensional IMSA function according to claim 1, characterized in that: The dual-frequency power amplifier digital predistortion model based on the two-dimensional IMSA function is derived from the 2-D-DPD model by the segmentation and approximation method; when segmentation and approximation are performed according to the square of the amplitude of the two input signals of the model, its expression is shown in formula (2): Among them, the subscripts i=1 and 2 represent the first frequency band and the second frequency band respectively. is the basis function of the model in each small area after segmentation; M is the memory depth, L1 and L2 represent the number of partitions in the |y1(n)| and |y2(n)| dimensions respectively. and They represent the partition thresholds on the corresponding dimensions and can be preset; and They respectively represent the three linear coefficients in each small area after segmentation when the memory depth is m; y1(nm) represents the digital baseband signal input to the first frequency band model at moment nm, and y2(nm) represents the digital baseband signal input to the second frequency band model at moment nm.
4. A dual-frequency power amplifier digital predistortion method based on a two-dimensional IMSA function according to any one of claims 1 to 3, characterized in that: The following steps are involved: 4.
1. The first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) are sent to a dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function. The dual-frequency power amplifier digital predistorter based on a two-dimensional IMSA function includes a dual-frequency power amplifier digital predistortion model based on a two-dimensional IMSA function. The first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) are predistorted to generate a first frequency band digital predistortion signal x pre1 (n) and the second frequency band digital predistortion signal x pre2 (n); 4.
2. Digital predistortion signal x in the first frequency band pre1 (n) performing digital-to-analog conversion and sending the converted signal to a first modulator to generate a radio frequency signal in a first frequency band; For the second frequency band digital predistortion signal x pre2 (n) performing digital-to-analog conversion and sending the result to a second modulator to generate a radio frequency signal in a second frequency band; 4.
3. Combining the first frequency band radio frequency signal and the second frequency band radio frequency signal into one signal and sending it to the power amplifier for amplification and output; 4.
4. The output signal of the power amplifier is attenuated and coupled, and then passes through the power divider to be divided into two equal signals; 4.
5. The first output signal of the power divider is demodulated, filtered, and converted digitally to analog by a first demodulator, a first low-pass filter, and a first analog-to-digital converter in sequence to generate a first-band output digital baseband signal y1(n); the second output signal of the power divider is demodulated, filtered, and converted digitally to analog by a second demodulator, a second low-pass filter, and a second analog-to-digital converter in sequence to generate a second-band output digital baseband signal y2(n); 4.
6. Perform dual-frequency power amplifier digital predistortion model training based on the two-dimensional IMSA function on the first frequency band input digital baseband signal x1(n), the second frequency band input digital baseband signal x2(n), the first frequency band output digital baseband signal y1(n), and the second frequency band output digital baseband signal y2(n) to determine the first frequency band predistortion parameters and the second frequency band predistortion parameters.
5. The method for digital predistortion of a dual-frequency power amplifier based on a two-dimensional IMSA function according to claim 4, characterized in that: In step 4.6, the training of the dual-frequency power amplifier digital pre-distortion model based on the two-dimensional IMSA function includes the following steps: 5.
1. Obtain a first frequency band input digital baseband signal x1(n) and a first frequency band output digital baseband signal y1(n) and perform normalization and alignment processing; obtain a second frequency band input digital baseband signal x2(n) and a second frequency band output digital baseband signal y2(n) and perform normalization and alignment processing; 5.
2. When the model is segmented and approximated according to the amplitudes of the two input signals, formula (1) is used as the digital pre-distortion model; when the model is segmented and approximated according to the squares of the amplitudes of the two input signals, formula (2) is used as the digital pre-distortion model; 5.
3. Determine the memory depth M, the number of partitions L1 and L2 of the basis function in two dimensions, and the pre-set partition threshold and 5.
4. Based on the memory depth M, the number of partitions of the basis function in two dimensions L1 and L2 and the pre-set partition threshold and The first frequency band output digital baseband signal y1(n) and the second frequency band output digital baseband signal y2(n) are used as model input, and the first frequency band input digital baseband signal x1(n) and the second frequency band input digital baseband signal x2(n) are used as model output to perform predistortion parameter training, and extract and determine the first frequency band predistortion parameters and the second frequency band predistortion coefficients.
6. The dual-frequency power amplifier digital predistortion method based on two-dimensional IMSA function according to claim 5, characterized in that: In step 5.4, the specific pre-distortion coefficient extraction process is The following steps are involved: 6.
1. According to the constraints of the input y1(n) and y2(n) of the digital pre-distortion model, find the model input samples y′1(n) and y′2(n) and the corresponding model output samples x′1(n) and x′2(n) belonging to the same small plane area; 6.
2. The relationship between the input and output samples of the model in each small plane area is written as: Where Y′ is the input matrix of the model, X′ is the corresponding model output vector, W′ is the model coefficient in each small area, N′ is the number of samples in each small area, and Q′=3M is the number of W′; 6.
3. Using the least squares (LS) algorithm, the corresponding solution is expressed as in represents the extracted predistortion coefficient, (·) H represents the conjugate transpose of a complex matrix, (·) -1 Represents the inverse of a complex matrix.
