CN113630091B - Power amplifier and predistortion model generation method and device thereof - Google Patents
Power amplifier and predistortion model generation method and device thereof Download PDFInfo
- Publication number
- CN113630091B CN113630091B CN202010390260.XA CN202010390260A CN113630091B CN 113630091 B CN113630091 B CN 113630091B CN 202010390260 A CN202010390260 A CN 202010390260A CN 113630091 B CN113630091 B CN 113630091B
- Authority
- CN
- China
- Prior art keywords
- model
- power amplifier
- predistortion
- determining
- coding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 79
- 239000011159 matrix material Substances 0.000 claims abstract description 74
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 72
- 230000002068 genetic effect Effects 0.000 claims abstract description 58
- 238000012216 screening Methods 0.000 claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 48
- 230000006870 function Effects 0.000 claims description 39
- 238000012545 processing Methods 0.000 claims description 22
- 238000011156 evaluation Methods 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 13
- 230000007246 mechanism Effects 0.000 claims description 13
- 230000035772 mutation Effects 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000003446 memory effect Effects 0.000 description 3
- 238000010187 selection method Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000005070 ripening Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F1/00—Details of amplifiers with only discharge tubes, only semiconductor devices or only unspecified devices as amplifying elements
- H03F1/32—Modifications of amplifiers to reduce non-linear distortion
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03F—AMPLIFIERS
- H03F3/00—Amplifiers with only discharge tubes or only semiconductor devices as amplifying elements
- H03F3/20—Power amplifiers, e.g. Class B amplifiers, Class C amplifiers
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Physics & Mathematics (AREA)
- Nonlinear Science (AREA)
- Amplifiers (AREA)
Abstract
The embodiment of the invention provides a power amplifier and a predistortion model generation method and device thereof, wherein the method comprises the following steps: screening the values of model parameters of a power amplifier model by using a genetic algorithm, and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model; model coefficients of the power amplifier model are determined based on the model input matrix and training data. According to the power amplifier and the predistortion model generation method and device thereof, key items influencing the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Description
Technical Field
The present invention relates to the field of communications technologies, and in particular, to a power amplifier and a predistortion model generating method and apparatus thereof.
Background
The efficiency and linearity of the rf power amplifier, which is the largest energy consuming device in the base station, directly affects the power consumption and the transmitted signal quality of the base station. In order to reduce the power consumption of the base station and improve the quality of the transmitted signal, digital Pre-distortion (DPD) technology is often used for correcting the linearity of the power amplifier.
In the prior art, the key of the DPD technology is to obtain a precise power amplifier model, a generalized memory polynomial (General Memory Polynomial, GMP) model is one of common models of the power amplifier model, and the GMP model has the advantages of high modeling precision, convenience for hardware implementation and the like, and is widely applied to base stations. Usually, based on training data, a least squares method is used for solving to obtain model coefficients of a GMP model, then a DPD model is determined based on polynomial inversion, and a signal is subjected to pre-distortion processing according to the DPD model, so that an output signal after cascading of power amplifiers is linearly amplified. In the process of determining the model coefficients of the GMP model, three model parameters (signal memory depth, model memory depth and nonlinear order) of the GMP model are continuously valued, the scale of a model input matrix of the GMP model is determined by the model parameters and the length N of training data, the number of rows of the model input matrix is (N+1), the number of columns is (I+1) · (J+1) · (K+1), wherein I is the maximum value of the signal memory depth, J is the maximum value of the model memory depth, and K is the maximum value of the nonlinear order.
However, as the ultra-high rate transmission requirement increases, the bandwidth of the modulation signal increases, and the ultra-wideband signal causes the memory effect of the power amplifier to be sharply increased, so that when modeling the strong memory power amplifier by using the GMP model, in order to accurately fit the power amplifier, the value of the model parameter used for representing the memory effect in the model needs to be increased. The memory depth of the GMP model is increased, the size of an input matrix of the model and the number of matrix conditions are increased, the stability of solving model coefficients is affected, the linear performance of DPD is reduced, the hardware resources occupied by DPD technology in a product are correspondingly increased, and the power consumption of the product is increased.
Disclosure of Invention
The embodiment of the invention provides a power amplifier and a predistortion model generation method and device thereof, which are used for solving the technical problems in the prior art.
In order to solve the above technical problems, in one aspect, an embodiment of the present invention provides a power amplifier and a predistortion model generating method thereof, including:
screening the values of model parameters of a power amplifier model by using a genetic algorithm, and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model;
model coefficients of the power amplifier model are determined based on the model input matrix and training data.
Further, the method for screening the values of the model parameters of the power amplifier model by using the genetic algorithm, and determining the model input matrix of the power amplifier model specifically includes:
screening the values of the model parameters by using a genetic algorithm, and determining key terms in a polynomial of the power amplifier model, wherein the key terms are the terms with the greatest influence on model precision in the polynomial of the power amplifier model;
and determining the model input matrix according to the key terms.
Further, the method for screening the values of the model parameters by using a genetic algorithm, and determining key terms in the polynomial of the power amplifier model specifically includes:
encoding the parameter value combinations of the model parameters, and determining the encoding value of each parameter value combination;
randomly selecting a preset number of coded values from all the coded values to serve as an initial group;
evaluating modeling precision by using a preset fitness function, and performing selection operation, crossover operation and mutation operation;
updating the iteration group until reaching a preset iteration ending condition, and outputting the coding value of the parameter value combination of the key item;
and determining the key item according to the coding value of the parameter value combination of the key item.
Further, the encoding the parameter value combination of the model parameters specifically includes:
performing decimal integer coding on the combination of the signal memory depth and the module memory depth in the model parameters;
and performing parity coding on nonlinear orders in the model parameters.
Further, the predetermined fitness function is determined based on a normalized mean square error of the signal.
Further, the evaluating modeling accuracy by using a preset fitness function specifically includes:
Stretching the preset fitness function;
and evaluating modeling accuracy by using the fitness function after the stretching treatment.
Further, the determining the model coefficient of the power amplifier model based on the model input matrix and training data specifically includes:
acquiring the training data, wherein the training data comprises an input signal and an output signal of a power amplifier;
and determining the model coefficient by adopting a least square method according to the model input matrix and the training data.
Further, after determining the model coefficients of the power amplifier model based on the model input matrix and training data, the method further includes:
and inverting the polynomial of the power amplifier model to determine a digital predistortion model.
Further, after the determining the digital predistortion model, the method further comprises:
and carrying out predistortion processing on the signal input into the power amplifier according to the digital predistortion model.
In another aspect, an embodiment of the present invention provides a power amplifier and a predistortion model generating apparatus thereof, including:
the screening module is used for screening the values of the model parameters of the power amplifier model by utilizing a genetic algorithm and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model;
And the generation module is used for determining the model coefficient of the power amplifier model based on the model input matrix and training data.
Further, the screening module includes a first screening sub-module and a second screening sub-module, wherein:
the first screening submodule is used for screening the values of the model parameters by utilizing a genetic algorithm, and determining key terms in a polynomial of the power amplifier model, wherein the key terms are terms with the greatest influence on model precision in the polynomial of the power amplifier model;
the second screening submodule is used for determining the model input matrix according to the key terms.
Further, the first screening submodule comprises a coding unit, an initializing unit, an evaluation mechanism unit, an output unit and a determining unit, wherein:
the coding unit is used for coding the parameter value combination of the model parameters and determining the coding value of each parameter value combination;
the initialization unit is used for randomly selecting a preset number of coded values from all the coded values to serve as an initial group;
the evaluation mechanism unit is used for evaluating modeling precision by utilizing a preset fitness function and performing selection operation, crossover operation and mutation operation;
The output unit is used for updating the iteration group until reaching a preset iteration ending condition, and outputting the coding value of the parameter value combination of the key item;
the determining unit is used for determining the key item according to the coding values of the parameter value combination of the key item.
Further, the coding unit comprises a first coding subunit and a second coding subunit, wherein:
the first coding subunit is used for performing decimal integer coding on the combination of the signal memory depth and the module memory depth in the model parameters;
the second encoding subunit is configured to perform parity encoding on the nonlinear order in the model parameters.
Further, the predetermined fitness function is determined based on a normalized mean square error of the signal.
Further, the evaluation mechanism unit comprises a stretching processing subunit and an evaluation subunit, wherein:
the stretching processing subunit is used for stretching the preset fitness function;
the evaluation subunit is used for evaluating modeling accuracy by using the fitness function after the stretching treatment.
Further, the generating module includes an obtaining sub-module and a determining sub-module, wherein:
The acquisition sub-module is used for acquiring the training data, wherein the training data comprises an input signal and an output signal of a power amplifier;
the determining submodule is used for determining the model coefficient by adopting a least square method according to the model input matrix and the training data.
Further, the method also comprises a digital predistortion model generation module;
the digital predistortion model generation module is used for inverting the polynomial of the power amplifier model to determine a digital predistortion model.
Further, the device also comprises a predistortion processing module;
the predistortion processing module is used for carrying out predistortion processing on signals input into the power amplifier according to the digital predistortion model.
In still another aspect, an embodiment of the present invention provides an electronic device, including: the computer program comprises a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor realizes the steps of the method provided in the first aspect when executing the computer program.
In yet another aspect, embodiments of the present invention provide a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method provided in the first aspect described above.
According to the power amplifier and the predistortion model generation method and device thereof, key items influencing the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Drawings
FIG. 1 is a diagram of a lookup table based on a lookup table in the prior art;
fig. 2 is a schematic diagram of a power amplifier and a predistortion model generation method according to an embodiment of the present invention;
FIG. 3 is a flowchart of the genetic algorithm provided by the embodiment of the invention applied to the screening of the power amplifier model;
fig. 4 is a schematic diagram of a power amplifier and a predistortion model generating device according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The efficiency and linearity of a radio frequency power amplifier (abbreviated as "power amplifier") as the largest energy consuming device in a base station directly affect the power consumption and the transmitted signal quality of the base station. In order to reduce the power consumption of the base station and improve the quality of the transmitted signal, DPD techniques are often used for correcting the linearity of the power amplifier. The DPD technology principle is that a processor which is reciprocal to a power amplifier behavior model (called as a power amplifier model for short) is established in a digital domain to carry out pre-distortion processing on signals, so that output signals after cascading power amplifiers are amplified linearly. The key of the DPD technology is to obtain an accurate power amplifier model, and the GMP model has the advantages of high modeling precision, convenience for hardware implementation and the like, and is widely applied to a base station.
The polynomial expression of the GMP model is as follows:
wherein y (n) is a power amplifier output signal, x (n) is a power amplifier input signal, I represents a signal memory depth, J represents a module memory depth, K represents a nonlinear order, I is a maximum value of the signal memory depth, J is a maximum value of the module memory depth, K is a maximum value of the nonlinear order, x (n-I) is a delay I of x (n) relative to x (n), and x (n-J) is a delay J of x (n) relative to x (n). b ijk The parameters are valued for coefficients combined into terms of ijk.
The matrix expression of the GMP model is as follows:
y＝X·b (2)
y＝[y(n)，y(n+1)，…，y(n+N)] T (3)
b＝[b 0，0，0 ，b 0，0，1 ，…，b i，j，k ，…，b I，J，K ] T (5)
Wherein y (n) is the power amplifier output signal, x (n) is the power amplifier input signal, I represents the signal memory depth, J represents the modulus memory depth, K represents the nonlinear order, I is the maximum value of the signal memory depth, J is the maximum value of the modulus memory depth, K is the maximum value of the nonlinear order, x (n-I) is the delay I of x (n) relative to x (n), x (n-J) is the delay J of x (n) relative to b ijk The parameters are combined into the coefficients of terms of ijk, N is the length of training data for training the power amplifier model, X is the model input matrix of the GMP model, b is the model coefficients of the GMP model, and y is the model output matrix of the GMP model.
The scale of the model input matrix X is determined by the length of the model parameters and training data, the number of the rows is (N+1), the number of the columns is (I+1) · (J+1) · (K+1), the number of the model coefficients b is equal to the number of the columns of the model input matrix X, and the number of the model coefficients b is (I+1) · (J+1) · (K+1).
On the one hand, based on training data, solving the formula (2) by adopting a least square method, and determining an expression of the model coefficient b as follows:
b＝(X H X) -1 X H y (6)
wherein, X is the model input matrix of the GMP model, b is the model coefficient of the GMP model, and y is the model output matrix of the GMP model.
After the model coefficient b of the GMP model is determined, a DPD model is determined based on polynomial inversion, and the signal is subjected to pre-distortion processing according to the DPD model, so that the output signal after cascading the power amplifier is linearly amplified.
On the other hand, when using the indirect learning structure to perform DPD modeling, only the left and right variables of the power amplifier model equation (1) need to be exchanged, that is, the matrix Y is constructed by using the power amplifier output signal Y (n), and the power amplifier input signal x (n) is used as the desired signal, the matrix expression of the DPD model is as follows:
x＝Y·c (7)
where c is the model coefficient of the DPD model.
Based on training data, solving the formula (7) by adopting a least square method, and determining the expression of the model coefficient c as follows:
c＝(Y H Y) -1 Y H x (8)
after determining the model coefficient c of the DPD model, the polynomial expression of the DPD model can be obtained as follows:
fig. 1 is a table lookup flow chart based on a lookup table in the prior art, as shown in fig. 1, DPD technology in a base station product is usually implemented in a field programmable gate array (Field Programmable Gate Array, FPGA) by adopting a table lookup mode, that is, multiplying predistortion coefficients and nonlinear terms of signal modulus values, adding the products, storing the products in a random access memory (Random Access Memory, RAM) in the FPGA after quantization, and performing table lookup according to signal amplitude. The number of DPD tables is determined by the memory parameters i and j of the GMP model, i.e. each combination of i and j corresponds to a table, and the data stored in the table is calculated according to the above formula (10), where AMP is the quantized signal amplitude. The larger the memory depth is, the more the number of the tables is, the more FPGA resources are correspondingly occupied, the stored data in the tables are obtained by multiplying and adding model coefficients and signal amplitudes, the model coefficients are obtained by establishing a model input matrix and solving by using a least square method, the higher the model complexity is, and the larger the condition number of the model input matrix is, the more unstable the coefficient solving is.
When the DPD technology is used in the base station, a corresponding model is usually selected according to the power amplification characteristics, and model parameters are adjusted according to the signal bandwidth after the model is selected. And when the signal bandwidth is increased, the memory effect of the power amplifier is enhanced, and the fitting precision of the model to the power amplifier characteristic is improved by increasing the value of the parameter memory depth representing the memory of the model. And when the memory depth of the model is increased, the matrix scale corresponding to the model is correspondingly increased, the matrix inversion complexity is increased when the matrix scale of the model is larger, and the condition number of the second matrix is increased to influence the stability of the coefficient solution. In addition, when the table look-up method is applied to realize the DPD technology, the number of the tables and the number of the multipliers directly depend on the model memory depth, and when the model memory depth is increased, the number of the tables is increased, and the FPGA resources occupied by the table storage and calculation are increased, so that the base station product is not beneficial to reducing the power consumption and the cost.
Taking the most applied GMP model of DPD technology as an example, if the maximum values of 3 model parameters of signal memory depth, module memory depth and nonlinear order of the GMP model are I, J, K respectively and are continuous values, the corresponding column number of the model input matrix is (I+1) · (J+1) · (K+1), the required number of form sheets (I+1) · (J+1) is realized by the model, and the required number of vector multipliers is (I+1). From the above analysis, when the memory parameters (signal memory depth i and module memory depth j) of the GMP model are increased, the FPGA resources consumed for implementation are drastically increased, which poses serious challenges for system power consumption and cost control.
In order to solve the technical problems, aiming at the application of ultra wideband signal DPD technology, the embodiment of the invention introduces a genetic algorithm to screen key memory items of the model when increasing the maximum memory depth of the GMP model, and reduces consumed hardware resources and power consumption on the premise of ensuring the modeling accuracy of the model.
Fig. 2 is a schematic diagram of a power amplifier and a predistortion model generation method thereof according to an embodiment of the present invention, and as shown in fig. 2, an embodiment of the present invention provides a power amplifier and a predistortion model generation method thereof, and an execution body thereof is a power amplifier and a predistortion model generation device thereof. The method comprises the following steps:
step S201, screening the values of model parameters of a power amplifier model by utilizing a genetic algorithm, and determining a model input matrix of the power amplifier model; the power amplifier model is a generalized memory polynomial model.
Specifically, the embodiment of the invention adopts a GMP model as a model of a power amplifier, and before the GMP model is generated, firstly, the genetic algorithm is utilized to screen the values of model parameters of the GMP model, and the model input matrix X of the optimized GMP model is determined.
Under the condition that the values of the parameters are not screened, the parameters are continuously valued, the expression of the model input matrix X is shown in the formula (4), the scale of the model input matrix X is determined by the model parameters and the length of training data, the number of lines is (N+1), and the number of columns is (I+1) · (J+1) · (K+1).
According to the embodiment of the invention, the genetic algorithm is utilized to screen the values of the model parameters of the GMP model, the values of the model parameters are not continuous after screening, only key items in the polynomial of the GMP model are reserved, redundant items are removed, the number of rows and/or the number of columns of the model input matrix X are reduced on the premise of ensuring modeling accuracy, and the complexity of the GMP model is reduced.
The key terms refer to terms with the greatest influence on model precision in the polynomial of the GMP model, and the number of the key terms can be configured according to actual conditions. Redundancy terms refer to terms of the polynomial of the GMP model that have less impact on model accuracy.
The model parameters of the GMP model comprise a signal memory depth I, a model memory depth J and a nonlinear order K, wherein I is the maximum value of the signal memory depth, J is the maximum value of the model memory depth, K is the maximum value of the nonlinear order, and I, J, K can be configured according to actual conditions.
Step S202, determining model coefficients of the power amplifier model based on the model input matrix and training data.
Specifically, after determining the model input matrix X of the optimized GMP model, the model coefficient b of the GMP model may be determined by solving the above formula (2) using a preset solution algorithm based on training data.
The solution algorithm may be a least squares algorithm, a machine learning algorithm, a linear regression algorithm, and the like.
According to the power amplifier and the predistortion model generation method thereof, provided by the embodiment of the invention, key items influencing the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, on the premise of ensuring the modeling precision, on one hand, the complexity of the model is reduced, namely, the matrix scale of the model is reduced, the condition number of the matrix is reduced, the calculation process is simplified, the calculation stability of the model coefficient is improved, on the other hand, the number of FPGA (field programmable gate array) tables and the number of vector multiplications required by hardware realization can be reduced, the consumption of the whole hardware is obviously reduced, and the power consumption and the cost control of a base station product are facilitated.
Based on any one of the foregoing embodiments, further, the filtering the values of the model parameters of the power amplifier model by using a genetic algorithm, to determine a model input matrix of the power amplifier model specifically includes:
Screening the values of the model parameters by using a genetic algorithm, and determining key terms in a polynomial of the power amplifier model, wherein the key terms are the terms with the greatest influence on model precision in the polynomial of the power amplifier model;
and determining the model input matrix according to the key terms.
Specifically, in the embodiment of the present invention, the genetic algorithm is used to screen the values of the model parameters of the GMP model, and the specific steps of determining the model input matrix of the optimized GMP model are as follows:
firstly, screening the values of model parameters of the GMP model by utilizing a genetic algorithm, and determining key terms in a polynomial of the GMP model.
The key terms refer to terms with the greatest influence on model precision in the polynomial of the GMP model, and the number of the key terms can be configured according to actual conditions.
Then, a model input matrix for the GMP model is determined from all key terms.
After screening, the values of the model parameters are not continuous any more, only key terms in polynomials of the GMP model are reserved, redundant terms are removed, and the model input matrix X of the optimized GMP model can be obtained. Redundancy terms refer to terms of the polynomial of the GMP model that have less impact on model accuracy.
On the premise of ensuring modeling accuracy, the number of rows and/or columns of the model input matrix X is reduced, and the complexity of the GMP model is reduced.
According to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Based on any one of the foregoing embodiments, further, the filtering the values of the model parameters by using a genetic algorithm, to determine key terms in the polynomial of the power amplifier model specifically includes:
encoding the parameter value combinations of the model parameters, and determining the encoding value of each parameter value combination;
randomly selecting a preset number of coded values from all the coded values to serve as an initial group;
evaluating modeling precision by using a preset fitness function, and performing selection operation, crossover operation and mutation operation;
updating the iteration group until reaching a preset iteration ending condition, and outputting the coding value of the parameter value combination of the key item;
And determining the key item according to the coding value of the parameter value combination of the key item.
Specifically, fig. 3 is an application flowchart of the genetic algorithm provided in the embodiment of the present invention in power amplifier model screening, as shown in fig. 3, in the embodiment of the present invention, when key items of a GMP model are screened by applying the genetic algorithm, it is first required to determine a screening initial model, then, based on the initial model, the screening item is added, and the key items of a final model are determined according to an evaluation mechanism of model accuracy.
Based on the GMP model, the polynomial expression of the GMP model is split into a memory term and a memory-free term according to whether the memory is included or not, and the expression is as follows:
wherein y (n) is a power amplifier output signal, x (n) is a power amplifier input signal, I represents a signal memory depth, J represents a module memory depth, K represents a nonlinear order, I is a maximum value of the signal memory depth, J is a maximum value of the module memory depth, K is a maximum value of the nonlinear order, x (n-I) is a delay I of x (n) relative to x (n), and x (n-J) is a delay J of x (n) relative to x (n). b ijk The parameters are valued for coefficients combined into terms of ijk. b 00k x(n)|x(n)| k For memory-less items, k is E [0, K]，b ijk x(n-i)|x(n-j)| k For memory items, i.e. [1, I ]]，j∈[1，J]，k∈[0，K]。
And taking the memory-free item as an initial model, taking the memory item as a candidate item, and screening the model according to the flow of a genetic algorithm. The genetic algorithm can be divided into coding and decoding design, initial group selection, evaluation mechanism, selection operation, crossover operation, mutation operation and other processes according to functions.
(1) And (5) encoding and decoding design.
And encoding the parameter value combinations of the model parameters, and determining the encoding value of each parameter value combination.
For example, the complete candidate after the encoding is completed is denoted by C, and the expression of C is as follows:
wherein c ij k odd Combination ijk for parameter value odd Coded value of c ij k even Combination ijk for parameter value even Is a coded value of (a).
The candidates are progressively coded in sequence starting from 1, the coding function is represented by code (), and the mapping relation of the candidate codes is as follows:
code(C)＝[1，2，3，…，D] (13)
where D is the total number of candidates.
The inverse of decoding into codes is represented by the function decode ().
(2) Initial population selection. Before the genetic algorithm iterates, the encoded values of the initial candidates need to be given, i.e. the initial population is selected. Assuming that M memory items are selected from candidate items to be used as final model combinations, after the coding mode is determined, M values are randomly selected from D codes to be used as codes D0 of the initial group in order to promote universality of the initial group, D 0 =rand (D, M), pair D 0 Performing decoding operation to obtain memory item C corresponding to the initial group 0 Is formulated as follows
C o ＝decode(D 0 ) (14)
(3) Evaluation mechanism. The evaluation mechanism corresponds to the fitness in the genetic algorithm, namely an index for evaluating the modeling precision of the screening model.
And evaluating modeling accuracy by using a preset fitness function, wherein the preset fitness function can be sum error vector amplitude and the like determined based on the normalized mean square error of the signal.
(4) And selecting operation. After the fitness function is determined, a certain criterion is selected as a basis for whether the candidate can be reserved in the next algorithm iteration.
The embodiment of the invention adopts proportion selection as an individual selection method, and the proportion selection, namely the probability that candidates are selected and reserved in the next algorithm iteration is in direct proportion to the fitness of the individual.
The proportion selection method comprises the following steps: first, the sum of fitness function values of all individuals in the population, Σe (X i ). Next, the proportion E (X) of the fitness function value of each individual is calculated i )/∑E(X i ) It is the probability that an individual is inherited into the next generation population. Finally, a ratio selection operator operation (i.e., generating a random number between 0 and 1 and comparing the size of the fitness ratio) is used to determine whether each individual is selected.
(5) Crossover operations and mutation operations. The cross operation is toCrossover probability P c Crossing and exchanging memory items in the selected model; the variation operation is based on the variation probability P m And extracting memory items from the unselected candidate items to replace the items in the selected model. The purpose of the crossover and mutation operation is to increase the diversity of the selected model and improve the modeling precision of the genetic algorithm screening model.
Crossover probability P c Probability of variation P m Can be configured according to actual conditions.
(6) Updating the iteration group until reaching the preset iteration ending condition, and outputting the coding value of the parameter value combination of the key items.
And finally, determining the key item according to the coding value of the parameter value combination of the key item. The items corresponding to the parameter value combinations screened by the genetic algorithm are used as key items, and the items corresponding to the parameter value combinations not screened are used as redundant items.
According to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Based on any one of the foregoing embodiments, further, the encoding the parameter value combination of the model parameter specifically includes:
performing decimal integer coding on the combination of the signal memory depth and the module memory depth in the model parameters;
and performing parity coding on nonlinear orders in the model parameters.
Specifically, in the embodiment of the invention, in the process of encoding the parameter value combination of the model parameters, in order to reduce the encoding quantity, the convergence speed of the algorithm is increased.
And performing decimal integer coding on the candidate item, namely the signal memory depth i and the modular value memory depth j in the memory item according to an increment rule.
Dividing nonlinear order k values into two groups k according to parity odd And k even Which is provided withIn (k) odd ＝[1，3，5，…]，k even ＝[0，2，4，…]Each signal memory depth i and the module memory depth j are combined to correspond to two coding valuesAnd->The complete candidate after the encoding is completed is denoted by C, which is expressed as follows:
wherein,combination ijk for parameter value odd Encoded value of->Combination ijk for parameter value even Is a coded value of (a).
The candidates are progressively coded in sequence starting from 1, the coding function is represented by code (), and the mapping relation of the candidate codes is as follows:
code(C)＝[1，2，3，…，D] (13)
where D is the total number of candidates, d=2ij.
The inverse of decoding into codes is represented by the function decode ().
According to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Based on any of the above embodiments, further, the preset fitness function is determined based on a normalized mean square error of the signal.
Specifically, in the embodiment of the invention, the modeling precision is characterized by adopting the Normalized Mean Square Error (NMSE) of the signal, and the expression of the NMSE is as follows:
wherein y (n) is the power amplifier output signal, and y' (n) is the power amplifier output signal estimated by the model. By NMSE 0 And separate NMSE i The normalized mean square error of the initial model (no memory model) and the algorithm ith iteration added memory term screening model are respectively represented, and the difference value of the initial model and the algorithm ith iteration added memory term screening model is used as algorithm fitness E, and then the expression of the fitness E is as follows:
E i ＝NMSE 0 -NMSE i (16)
according to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Based on any one of the foregoing embodiments, further, the evaluating modeling accuracy by using a preset fitness function specifically includes:
stretching the preset fitness function;
and evaluating modeling accuracy by using the fitness function after the stretching treatment.
Specifically, in the embodiment of the invention, when a proportion selection method is adopted for selection operation, a genetic algorithm converges in advance, namely, the phenomenon of 'early ripening', and the modeling precision of a final screening model is affected. According to the embodiment of the invention, the retention probability of the model with higher precision is increased by stretching the fitness of the selected model, so that the 'early ripening' phenomenon of a genetic algorithm is avoided. The fitness stretching algorithm is formulated as follows:
T＝T 0 (0.99 (g-1) ) (18)
Wherein E is i Representing the fitness of the ith selected model, L represents the number of selected models of each algorithm iteration, g represents the algorithm iteration number, T represents the stretching degree and T 0 Indicating the initial degree of stretch (default value 1).
According to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Based on any one of the foregoing embodiments, further, the determining, based on the model input matrix and training data, model coefficients of the power amplifier model specifically includes:
acquiring the training data, wherein the training data comprises an input signal and an output signal of a power amplifier;
and determining the model coefficient by adopting a least square method according to the model input matrix and the training data.
Specifically, in the embodiment of the present invention, based on the model input matrix and the training data, the specific steps for determining the model coefficients of the GMP model are as follows:
first, training data is acquired, the training data comprising an input signal and an output signal of a power amplifier.
The length of the training data can be configured according to the actual situation.
Then, based on the model input matrix and training data, the model coefficients of the GMP model are determined using a least squares method.
According to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Based on any of the foregoing embodiments, further, after determining the model coefficients of the power amplifier model based on the model input matrix and training data, the method further includes:
and inverting the polynomial of the power amplifier model to determine a digital predistortion model.
Specifically, in the embodiment of the present invention, after determining the model coefficients of the GMP model, the polynomial expression of the GMP model is determined.
And then inverting according to a polynomial to determine the digital predistortion model.
According to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Based on any of the foregoing embodiments, further comprising, after determining the digital predistortion model:
and carrying out predistortion processing on the signal input into the power amplifier according to the digital predistortion model.
Specifically, in the embodiment of the invention, after the digital predistortion model is determined, the digital predistortion model can be utilized to carry out predistortion processing on the signal input to the power amplifier, thereby eliminating nonlinear distortion of the power amplifier.
According to the power amplifier and the predistortion model generation method thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of guaranteeing the modeling precision.
Fig. 4 is a schematic diagram of a power amplifier and a predistortion model generating device thereof according to an embodiment of the present invention, and as shown in fig. 4, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, including: a screening module 401 and a generating module 402, wherein:
the screening module 401 is configured to screen values of model parameters of a power amplifier model by using a genetic algorithm, and determine a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model; the generation module 402 is configured to determine model coefficients of the power amplifier model based on the model input matrix and training data.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any of the above embodiments, further, the screening module includes a first screening sub-module and a second screening sub-module, wherein:
the first screening submodule is used for screening the values of the model parameters by utilizing a genetic algorithm, and determining key terms in a polynomial of the power amplifier model, wherein the key terms are terms with the greatest influence on model precision in the polynomial of the power amplifier model;
the second screening submodule is used for determining the model input matrix according to the key terms.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any one of the above embodiments, further, the first screening submodule includes an encoding unit, an initializing unit, an evaluation mechanism unit, an output unit, and a determining unit, where:
the coding unit is used for coding the parameter value combination of the model parameters and determining the coding value of each parameter value combination;
the initialization unit is used for randomly selecting a preset number of coded values from all the coded values to serve as an initial group;
the evaluation mechanism unit is used for evaluating modeling precision by utilizing a preset fitness function and performing selection operation, crossover operation and mutation operation;
The output unit is used for updating the iteration group until reaching a preset iteration ending condition, and outputting the coding value of the parameter value combination of the key item;
the determining unit is used for determining the key item according to the coding values of the parameter value combination of the key item.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any of the above embodiments, further, the coding unit includes a first coding subunit and a second coding subunit, wherein:
the first coding subunit is used for performing decimal integer coding on the combination of the signal memory depth and the module memory depth in the model parameters;
The second encoding subunit is configured to perform parity encoding on the nonlinear order in the model parameters.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any of the above embodiments, further, the preset fitness function is determined based on a normalized mean square error of the signal.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any of the above embodiments, further, the evaluation mechanism unit includes a stretch processing subunit and an evaluation subunit, wherein:
the stretching processing subunit is used for stretching the preset fitness function;
the evaluation subunit is used for evaluating modeling accuracy by using the fitness function after the stretching treatment.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any of the above embodiments, further, the generating module includes an acquiring sub-module and a determining sub-module, wherein:
the acquisition sub-module is used for acquiring the training data, wherein the training data comprises an input signal and an output signal of a power amplifier;
the determining submodule is used for determining the model coefficient by adopting a least square method according to the model input matrix and the training data.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any one of the above embodiments, further comprising a digital predistortion model generation module;
The digital predistortion model generation module is used for inverting the polynomial of the power amplifier model to determine a digital predistortion model.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Based on any of the above embodiments, further comprising a predistortion processing module;
the predistortion processing module is used for carrying out predistortion processing on signals input into the power amplifier according to the digital predistortion model.
Specifically, an embodiment of the present invention provides a power amplifier and a predistortion model generating device thereof, which are configured to execute a method in the foregoing corresponding embodiment, and specific steps of executing the method in the foregoing corresponding embodiment by using the device provided by the present embodiment are the same as those of the foregoing corresponding embodiment, and are not repeated herein.
According to the power amplifier and the predistortion model generation device thereof, key items affecting the modeling precision of the power amplifier in the generalized memory polynomial model are screened by utilizing a genetic algorithm, redundant items are removed, the complexity of the model is reduced, hardware resources are saved, and the power consumption is reduced on the premise of ensuring the modeling precision.
Fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, as shown in fig. 5, where the electronic device includes: a processor (processor) 501, a communication interface (Communications Interface) 502, a memory (memory) 503 and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 communicate with each other via the communication bus 504. The processor 501 may call a computer program stored on the memory 503 and executable on the processor 501 to perform the steps of:
screening the values of model parameters of a power amplifier model by using a genetic algorithm, and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model;
model coefficients of the power amplifier model are determined based on the model input matrix and training data.
Further, the logic instructions in the memory 503 described above may be implemented in the form of software functional units and may be stored in a computer readable storage medium when sold or used as a stand alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Further, embodiments of the present invention provide a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, are capable of performing the steps of the method embodiments described above, for example comprising:
Screening the values of model parameters of a power amplifier model by using a genetic algorithm, and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model;
model coefficients of the power amplifier model are determined based on the model input matrix and training data.
Further, embodiments of the present invention provide a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method embodiments described above, for example, including:
screening the values of model parameters of a power amplifier model by using a genetic algorithm, and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model;
model coefficients of the power amplifier model are determined based on the model input matrix and training data.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (18)
1. A power amplifier and a predistortion model generation method thereof are characterized by comprising the following steps:
screening the values of model parameters of a power amplifier model by using a genetic algorithm, and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model; the nonlinear order in the model parameters is coded in a parity coding mode; the modeling accuracy of the power amplifier model is obtained by using a preset fitness function; the preset fitness function is determined according to the difference value between the normalized mean square error of the signal corresponding to the initial power amplifier model and the normalized mean square error of the signal corresponding to the power amplifier model under the current iteration number;
model coefficients of the power amplifier model are determined based on the model input matrix and training data.
2. The power amplifier and the predistortion model generation method according to claim 1, wherein the filtering the values of the model parameters of the power amplifier model by using a genetic algorithm, determining the model input matrix of the power amplifier model, specifically comprises:
Screening the values of the model parameters by using a genetic algorithm, and determining key terms in a polynomial of the power amplifier model, wherein the key terms are the terms with the greatest influence on model precision in the polynomial of the power amplifier model;
and determining the model input matrix according to the key terms.
3. The power amplifier and the predistortion model generation method according to claim 2, wherein the filtering the values of the model parameters by using a genetic algorithm, determining key terms in the polynomial of the power amplifier model, specifically comprises:
encoding the parameter value combinations of the model parameters, and determining the encoding value of each parameter value combination;
randomly selecting a preset number of coded values from all the coded values to serve as an initial group;
evaluating modeling precision by using a preset fitness function, and performing selection operation, crossover operation and mutation operation;
updating the iteration group until reaching a preset iteration ending condition, and outputting the coding value of the parameter value combination of the key item;
and determining the key item according to the coding value of the parameter value combination of the key item.
4. The power amplifier and predistortion model generation method according to claim 3, wherein said encoding said parameter value combination of model parameters specifically comprises:
Performing decimal integer coding on the combination of the signal memory depth and the module memory depth in the model parameters;
and performing parity coding on nonlinear orders in the model parameters.
5. The power amplifier and the predistortion model generation method according to claim 3, wherein said evaluating modeling accuracy using a preset fitness function specifically comprises:
stretching the preset fitness function;
and evaluating modeling accuracy by using the fitness function after the stretching treatment.
6. The power amplifier and the predistortion model generation method thereof according to any one of claims 1-5, wherein said determining model coefficients of said power amplifier model based on said model input matrix and training data, specifically comprises:
acquiring the training data, wherein the training data comprises an input signal and an output signal of a power amplifier;
and determining the model coefficient by adopting a least square method according to the model input matrix and the training data.
7. The power amplifier and predistortion model generation method according to any one of claims 1-5, characterized in that after determining model coefficients of the power amplifier model based on the model input matrix and training data, further comprising:
A polynomial to the power amplifier model.
8. The power amplifier and predistortion model generation method according to claim 7, further comprising, after determining the digital predistortion model:
and carrying out predistortion processing on the signal input into the power amplifier according to the digital predistortion model.
9. A power amplifier and predistortion model generation apparatus thereof, comprising:
the screening module is used for screening the values of the model parameters of the power amplifier model by utilizing a genetic algorithm and determining a model input matrix of the power amplifier model; wherein the power amplifier model is a generalized memory polynomial model; the nonlinear order in the model parameters is coded in a parity coding mode; the modeling accuracy of the power amplifier model is obtained by using a preset fitness function; the preset fitness function is determined according to the difference value between the normalized mean square error of the signal corresponding to the initial power amplifier model and the normalized mean square error of the signal corresponding to the power amplifier model under the current iteration number;
and the generation module is used for determining the model coefficient of the power amplifier model based on the model input matrix and training data.
10. The power amplifier and predistortion model generation apparatus of claim 9, wherein said screening module comprises a first screening sub-module and a second screening sub-module, wherein:
the first screening submodule is used for screening the values of the model parameters by utilizing a genetic algorithm, and determining key terms in a polynomial of the power amplifier model, wherein the key terms are terms with the greatest influence on model precision in the polynomial of the power amplifier model;
the second screening submodule is used for determining the model input matrix according to the key terms.
11. The power amplifier and predistortion model generation apparatus according to claim 10, wherein said first screening submodule comprises an encoding unit, an initializing unit, an evaluation mechanism unit, an output unit, and a determining unit, wherein:
the coding unit is used for coding the parameter value combination of the model parameters and determining the coding value of each parameter value combination;
the initialization unit is used for randomly selecting a preset number of coded values from all the coded values to serve as an initial group;
the evaluation mechanism unit is used for evaluating modeling precision by utilizing a preset fitness function and performing selection operation, crossover operation and mutation operation;
The output unit is used for updating the iteration group until reaching a preset iteration ending condition, and outputting the coding value of the parameter value combination of the key item;
the determining unit is used for determining the key item according to the coding values of the parameter value combination of the key item.
12. The power amplifier and predistortion model generation apparatus of claim 11, wherein said coding unit comprises a first coding subunit and a second coding subunit, wherein:
the first coding subunit is used for performing decimal integer coding on the combination of the signal memory depth and the module memory depth in the model parameters;
the second encoding subunit is configured to perform parity encoding on the nonlinear order in the model parameters.
13. The power amplifier and predistortion model generation apparatus according to claim 11, wherein said evaluation mechanism unit comprises a stretching processing subunit and an evaluation subunit, wherein:
the stretching processing subunit is used for stretching the preset fitness function;
the evaluation subunit is used for evaluating modeling accuracy by using the fitness function after the stretching treatment.
14. The power amplifier and predistortion model generation apparatus according to any one of claims 9-13, wherein said generation module comprises an acquisition sub-module and a determination sub-module, wherein:
The acquisition sub-module is used for acquiring the training data, wherein the training data comprises an input signal and an output signal of a power amplifier;
the determining submodule is used for determining the model coefficient by adopting a least square method according to the model input matrix and the training data.
15. The power amplifier and predistortion model generation apparatus of any one of claims 9 to 13, further comprising a digital predistortion model generation module;
the digital predistortion model generation module is used for inverting the polynomial of the power amplifier model to determine a digital predistortion model.
16. The power amplifier and predistortion model generation apparatus of claim 15, further comprising a predistortion processing module;
the predistortion processing module is used for carrying out predistortion processing on signals input into the power amplifier according to the digital predistortion model.
17. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the steps of the power amplifier and the predistortion model generation method of any of claims 1 to 8.
18. A non-transitory computer readable storage medium having stored thereon a computer program, characterized in that the steps of the power amplifier and the predistortion model generation method thereof according to any of claims 1 to 8 are implemented when said computer program is executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010390260.XA CN113630091B (en) | 2020-05-08 | 2020-05-08 | Power amplifier and predistortion model generation method and device thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010390260.XA CN113630091B (en) | 2020-05-08 | 2020-05-08 | Power amplifier and predistortion model generation method and device thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113630091A CN113630091A (en) | 2021-11-09 |
| CN113630091B true CN113630091B (en) | 2024-03-29 |
Family
ID=78377671
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010390260.XA Active CN113630091B (en) | 2020-05-08 | 2020-05-08 | Power amplifier and predistortion model generation method and device thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113630091B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023173383A1 (en) * | 2022-03-17 | 2023-09-21 | 华为技术有限公司 | Adaptive power control method, adaptive power processing system, and related device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789920A (en) * | 2009-12-29 | 2010-07-28 | 北京北方烽火科技有限公司 | Method and system for realizing self-adaptive predistortion power amplifier linearization |
| CN106411271A (en) * | 2016-08-10 | 2017-02-15 | 成都芯通科技股份有限公司 | Predistortion device for power amplifier and parameter selection method thereof |
| CN107276546A (en) * | 2016-04-08 | 2017-10-20 | 大唐移动通信设备有限公司 | A kind of digital pre-distortion processing method and device |
| CN110601665A (en) * | 2019-08-23 | 2019-12-20 | 海南电网有限责任公司 | Digital predistorter design method and device based on power amplifier model clipping |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9214969B2 (en) * | 2013-05-09 | 2015-12-15 | King Fahd University Of Petroleum And Minerals | Scalable digital predistortion system |
-
2020
- 2020-05-08 CN CN202010390260.XA patent/CN113630091B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789920A (en) * | 2009-12-29 | 2010-07-28 | 北京北方烽火科技有限公司 | Method and system for realizing self-adaptive predistortion power amplifier linearization |
| CN107276546A (en) * | 2016-04-08 | 2017-10-20 | 大唐移动通信设备有限公司 | A kind of digital pre-distortion processing method and device |
| CN106411271A (en) * | 2016-08-10 | 2017-02-15 | 成都芯通科技股份有限公司 | Predistortion device for power amplifier and parameter selection method thereof |
| CN110601665A (en) * | 2019-08-23 | 2019-12-20 | 海南电网有限责任公司 | Digital predistorter design method and device based on power amplifier model clipping |
Non-Patent Citations (3)
| Title |
|---|
| 一种优化的神经网络数字预失真方法;张烈等;西北工业大学学报;第32卷(第06期);全文 * |
| 基于遗传算法的功放预失真器模型研究与实现;佘小莉;;陕西电力;第43卷(第09期);全文 * |
| 张烈等.一种优化的神经网络数字预失真方法.西北工业大学学报.2014,第32卷(第06期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113630091A (en) | 2021-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103685111B (en) | Calculating method of digital pre-distortion parameters and pre-distortion system | |
| JP7451720B2 (en) | Predistortion methods, systems, devices and storage media | |
| WO2011139858A2 (en) | Nonlinear identification using compressed sensing and minimal system sampling | |
| EP2641325A1 (en) | Orthogonal basis function set for ditigal predistorter | |
| CN111985523A (en) | Knowledge distillation training-based 2-exponential power deep neural network quantification method | |
| CN113630091B (en) | Power amplifier and predistortion model generation method and device thereof | |
| CN109255429B (en) | Parameter decompression method for sparse neural network model | |
| CN113037226B (en) | Digital predistortion design method and device based on adaptive step length clipping method | |
| CN104796364A (en) | Pre-distortion parameter solving method and pre-distortion processing system | |
| CN111988046A (en) | Nonlinear system correction method and device and electronic equipment | |
| RU2676017C1 (en) | Device and method for adaptive linearization of analog radio path through dual-digital corrector | |
| CN115037340B (en) | Signal detection method, device, electronic equipment and storage medium | |
| CN113055323B (en) | Method and system for digital predistortion processing of communication system | |
| CN102522957B (en) | A kind of method improving predistortion performance of radio-frequency power amplifier | |
| CN108574649B (en) | Method and device for determining digital predistortion coefficient | |
| CN113516170B (en) | Image classification method based on Bayesian neural network random addition saturated structure | |
| CN103731106B (en) | A kind of segmentation digital pre-distortion method of radio-frequency (RF) power amplification | |
| CN114978194A (en) | Structure optimization method and device of original pattern LDPC code suitable for lossy source coding | |
| Huang et al. | The robust kernel conjugate gradient least mean p-power algorithm | |
| JP2019192125A (en) | Calculation device, calculation method, and program | |
| CN116993629B (en) | Smoothing method and device based on image decomposition, electronic equipment and storage medium | |
| CN113872614B (en) | Deep neural network-based Reed-Solomon code decoding method and system | |
| EP1482637A1 (en) | Method of influencing an input signal and predistorter | |
| CN107547053B (en) | Signal digital predistortion processing method and device | |
| JP7418619B2 (en) | Learning device for distortion compensation circuit and distortion compensation circuit |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |