CN111245375A - Power amplifier digital predistortion method of complex value full-connection recurrent neural network model - Google Patents

Power amplifier digital predistortion method of complex value full-connection recurrent neural network model Download PDF

Info

Publication number
CN111245375A
CN111245375A CN202010059771.3A CN202010059771A CN111245375A CN 111245375 A CN111245375 A CN 111245375A CN 202010059771 A CN202010059771 A CN 202010059771A CN 111245375 A CN111245375 A CN 111245375A
Authority
CN
China
Prior art keywords
power amplifier
model
neural network
output
complex
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.)
Granted
Application number
CN202010059771.3A
Other languages
Chinese (zh)
Other versions
CN111245375B (en
Inventor
党妮
徐常志
汤昊
靳一
汪滴珠
左金钟
杨丽
李明玉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Institute of Space Radio Technology
Original Assignee
Xian Institute of Space Radio Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Institute of Space Radio Technology filed Critical Xian Institute of Space Radio Technology
Priority to CN202010059771.3A priority Critical patent/CN111245375B/en
Publication of CN111245375A publication Critical patent/CN111245375A/en
Application granted granted Critical
Publication of CN111245375B publication Critical patent/CN111245375B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • H03F3/189—High frequency amplifiers, e.g. radio frequency amplifiers
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

A power amplifier digital predistortion method of a complex value full-connection recurrent neural network model is characterized in that the complex power amplifier model is simulated by utilizing the complex value full-connection recurrent neural network model, so that a power amplifier inverse model is obtained, and self-adaptive digital predistortion is realized. The power amplifier model is based on a complex value neural network theory, an improved complex value real-time recursive learning algorithm is adopted on the basis of the real-time recursive learning algorithm, and more accurate model approximation is realized for a power amplifier model at a transmitting end of a digital communication system. The invention combines the real-time recursive learning algorithm in the recurrent neural network, provides a complex value full-connection recurrent neural network model with better effect on the original real value recurrent neural network model, and further popularizes the complex value real-time recursive learning algorithm. Through simulation verification, the model structure and the algorithm have good performance in the aspects of training time and modeling accuracy, and can ensure high fitting degree of power amplifier nonlinearity.

Description

Power amplifier digital predistortion method of complex value full-connection recurrent neural network model
Technical Field
The invention belongs to the technical field of digital signal processing of a wireless communication system, and particularly relates to a digital predistortion method based on complex value neural network power amplifier modeling.
Background
In modern communication systems, due to the dual pressure of high-rate data transmission requirements and limited spectrum resources, Modulation schemes such as Quadrature Amplitude Modulation (QAM), Quadrature Phase Shift Keying (QPSK), Orthogonal Frequency Division Multiplexing (OFDM) are increasingly used in communication systems to improve spectrum utilization. However, such modulation techniques increase the design difficulty of the radio frequency power amplifier, and such signals are envelope modulation signals, have the characteristic of high Peak-to-average power ratio (PAPR), inevitably introduce nonlinear distortion, and under the same average power level, signals with higher PAPR are more sensitive to power amplifier nonlinearity, resulting in nonlinear increase. Moreover, many devices in the communication system have inherent nonlinearity, and when the envelope modulation signal passes through these devices, harmonic components and intermodulation distortion are generated, which causes nonlinearity, interference to adjacent channels and influences the performance of the communication system, so that the linearization of the power amplifier is a significant problem in modern communication. A common Power amplifier linearization technique is Power Back off (Power Back off), which is based on the principle that a Power amplifier is far away from a saturation region when operating, and operates in a linear region by means of Back off, although the Power Back off method is simple to implement, the operating efficiency is too low, and among various linearization techniques, digital predistortion is regarded as a most promising Power amplifier linearization technique by the industry due to the advantages of good linearity, wide bandwidth, high efficiency, full adaptivity, and the like.
In recent years, because the application of the Neural network is wider and wider, and the Neural network has good nonlinear system approximation capability, stronger learning capability, better robustness and self-adaption capability, the attention of the field of power amplifier modeling is attracted, an Artificial Neural Network (ANN) is used as one of modeling and predistortion technologies of a power amplifier and a transmitter for research, and more theories and practices are provided in the aspect of power amplifier modeling. The fusion of the Neural Network technology and the power amplifier predistortion technology benefits from the strong approximation effect of the Neural Network on nonlinear system modeling, the application field of a Complex-Valued Neural Network (CVNN) model is more and more extensive, but the application of the Complex-Valued Neural Network on the power amplifier behavior model modeling and the digital predistortion technology is still deficient.
Disclosure of Invention
The technical problem solved by the invention is as follows: aiming at the problems of insufficient model prediction accuracy, high calculation complexity, insufficient predistortion correction capability and the like of the existing power amplifier modeling method based on the neural network, the power amplifier digital predistortion method of the complex value full-connection recurrent neural network model is provided on the basis of the traditional real value neural network model and the training algorithm, a power amplifier behavior model can be established more accurately, a power amplifier nonlinear curve can be fitted well, namely, the model accuracy is good, and on the basis of high model accuracy, a power amplifier model inversion method is adopted, so that a predistorter with good prediction effect can be obtained.
The technical solution of the invention is as follows: a power amplifier digital predistortion method of a complex value full-connection recurrent neural network model comprises the following steps:
(1) sending signal data x (n) to a hardware communication system, and acquiring an output signal y (n) of the radio frequency power amplifier through a feedback channel of the hardware communication system;
(2) performing auto-correlation synchronization between y (n) and x (n), and performing synchronous alignment processing on the input and output signals;
(3) after normalization processing is carried out on x (n) and y (n), a complex value full-connection recurrent neural network model is used for carrying out preliminary power amplifier modeling;
(4) updating parameters of the established initial power amplifier model by using a recursive learning training algorithm to obtain a final power amplifier model;
(5) by utilizing an inversion method, a power amplifier output signal is used as a model input, a power amplifier input signal is used as a reference signal for modeling, and an inverse model of the power amplifier is obtained, namely the inverse model of the power amplifier is a digital predistorter of the power amplifier;
(6) an input signal x (n) enters a power amplifier digital predistorter to obtain an output sequence signal z (n), and the output sequence signal z (n) is processed by a power amplifier model to obtain an output sampling signal v (n);
(7) obtaining absolute error signals | e (n) | according to e (n) ═ x (n) — v (n), and judging the effect of the predistorter according to the size of | e (n) |.
In the complex value full-connected recurrent neural network model, the input, the weight and the output are complex numbers, and the complex value full-connected recurrent neural network model comprises N neurons, p external inputs and N feedback connecting lines, has a two-layer structure and is respectively an external input feedback layer and an output processing layer; in the network, the complex value of each neuron at the k-th time is output by yl(k) N, the external input is represented by p delay terms of s (k), which is a (1 × p) vector, where s (k) is the input sample sequence, and the total input vector p (k) of the entire network is represented by the output vector yl(k) S (k) and an offset input (1+ j) in series, as:
Figure BDA0002374066980000031
in the formula (I), the compound is shown in the specification,
Figure BDA0002374066980000032
(·)Ttranspose of the representation vector, (.)rAnd (·)iRepresenting the real and imaginary parts of a complex value or complex vector; for the l-th neuron, its weight is a vector of (p + N +1) × 1 dimensions, the same as the input vector dimensions, expressed as:
Figure BDA0002374066980000033
the complex matrix of the entire network is expressed by the formula: w ═ W1,...,wN]；
The output of each neuron is defined as: y isl(k)＝Φ(netl(k) Φ represents the complex-valued nonlinear activation function of the neuron, netl(k) Is the input of the activation function at time k, i.e. the linear sum of all the inputs of a certain node after applying weights, and is expressed as:
Figure BDA0002374066980000034
yl(k)＝Φr(netl(k))+jΦi(netl(k))＝ul(k)+jvl(k)。
the activation function is an initial transcendental function Φ (z) ═ tanh (z).
In the step (4), a recursive learning training algorithm is used for parameter updating, and the method specifically comprises the following steps:
step 1, supposing that the output of each neuron in the output layer is yt(k) N, with an output error of et(k) Error is represented by real part et r(k) And imaginary part et i(k) Composition, expressed as: e.g. of the typet(k)＝d(k)-yt(k)＝et r(k)+jet i(k)，et r(k)＝dr(k)-ut(k),et i(k)＝di(k)-vt(k)，d(k)＝dr(k)+jdi(k) Represents a reference signal;
step 2, taking the difference value of the reference signal and the output signal of the neuron as a cost function and expressing the difference value as
Figure BDA0002374066980000035
Step 3, weighting each weight w in the networkl,nAn update equation for e W, l 1., N1., p + N +1 is expressed as
Figure BDA0002374066980000041
η is the learning rate at which the user can learn,
Figure BDA0002374066980000042
to calculate the gradient;
Figure BDA0002374066980000043
is the conjugation of the complex number;
Figure BDA0002374066980000044
Figure BDA0002374066980000045
are respectively as
Figure BDA0002374066980000046
Figure BDA0002374066980000047
And
Figure BDA0002374066980000048
a simplified representation of;
Figure BDA0002374066980000049
δlnis a kronecker function.
Compared with the prior art, the invention has the advantages that:
(1) the method of the invention uses the FCRNN neural network to model the power amplifier, can express the nonlinear characteristic and the memory effect of the power amplifier rapidly and better, and has strong approaching ability to the power amplifier; the CRTRL (complex-valued real time learning) algorithm is used for training and learning the model, so that the power amplifier model can be accurately represented, and the linear correction is performed on the power amplifier; meanwhile, the CRTRL algorithm can enable the model to be converged more quickly, so that the training time is reduced, and the calculation complexity of the model is reduced;
(2) compared with the traditional power amplifier digital predistortion technology, the method has higher modeling accuracy due to the consideration of the influence of complex signals;
(3) the method can better model the power amplifier and adopt a direct learning type structure, so that the linear correction and compensation can be better carried out on the power amplifier.
(4) The method is not only suitable for establishing the predistortion model of the F-type power amplifier, but also suitable for other power amplifiers.
Drawings
FIG. 1 is a schematic diagram of a complex-valued fully-connected recurrent neural network architecture of the present invention;
FIG. 2 is a schematic diagram of an implementation of the method of the present invention;
FIG. 3 is a diagram illustrating the modeling effect of the FCRNN model according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating simulation of the effect of the predistorter built by the D-FCRNN according to the embodiment of the present invention.
Detailed Description
The invention adopts a complex connected recurrent neural network (FCRNN) to establish an effective power amplifier model, trains the model by using a complex real-time recurrent learning (CRTRL) algorithm, and obtains a predistorter model by inverting the trained model, which comprises the following steps:
step A, sending signal data x (n) to a hardware communication system, acquiring an output signal y (n) of a radio frequency power amplifier through a hardware feedback channel, and then entering step B;
b, performing autocorrelation synchronization according to the collected output signals y (n) and the input signals x (n), performing synchronization alignment processing on the input and output signals, and then entering the step C;
and C, normalizing the input signal x (n) and the sampled output signal y (n), and then performing primary power amplifier modeling by using a complex value full-connection recurrent neural network model.
The model of the fully-connected recurrent neural network is shown in fig. 1, and in the model structure of the complex value fully-connected recurrent neural network in fig. 1, the input, weight and output of the network are complex numbers, and the network structure is directly expanded from a real number field to a complex number field. Fig. 1 consists of N neurons (shown as circles in the figure), p external inputs and N feedback connections. The network has a two-layer structure, namely an external input feedback layer and an output processing layer. In the network, the complex value of each neuron at the k-th time is output by yl(k) Watch (A)I 1.. N. The external input is represented by p delay terms of s (k), which is a (1 × p) vector, where s (k) is the input sample sequence. So that the total input vector P (k) of the entire network is derived from the output vector yl(k) S (k) and a bias input (1+ j) connected in series, the bias input being used to introduce an external complex vector, the overall input expression of which can be described as follows:
Figure BDA0002374066980000051
in the formula (I), the compound is shown in the specification,
Figure BDA0002374066980000052
(·)Ttranspose of the representation vector, (.)rAnd (·)iRepresenting the real and imaginary parts of a complex value or complex vector.
For the l-th neuron, its weight is a vector of (p + N +1) × 1 dimensions (which may take a random value less than 1) the same as the input vector dimensions, expressed as:
Figure BDA0002374066980000053
the complex matrix of the whole neural network is expressed by the following formula: w ═ W1,...,wN]。
The output of each neuron is defined as: y isl(k)＝Φ(netl(k) Φ represents a complex nonlinear activation function of a neuron, and an initial transcendental function Φ (z) ═ tanh (z) is selected as the activation function in order to realize complex gradient learning. Netl(k) Is the input of the activation function at time k, i.e. the linear sum of all the inputs of a certain node after applying weights, and is expressed as:
Figure BDA0002374066980000061
for simplicity, y may be takenl(k) Expressed as:
yl(k)＝Φr(netl(k))+jΦi(netl(k))＝ul(k)+jvl(k)
and D, updating parameters of the established initial power amplifier model by using an improved RTRL training algorithm (called CRTRL in the invention) to obtain a final power amplifier model.
The CRTRL algorithm specifically comprises the following steps:
step 1, outputting each neuron in an output layer as yt(k) N, with an output error of et(k) Error is represented by real part et r(k) And imaginary part et i(k) Composition, expressed as: e.g. of the typet(k)＝d(k)-yt(k)＝et r(k)+jet i(k)，et r(k)＝dr(k)-ut(k),et i(k)＝di(k)-vt(k)。
Step 2, u in step 1t(k) And vt(k) D (k) d, i.e. the real and imaginary parts of the output, respectivelyr(k)+jdi(k) And denotes a learning signal, i.e., a reference signal. The cost function is the difference between the reference signal and the neuron output signal and is expressed as
Figure BDA0002374066980000062
Step 3, each weight w in the networkl,nAn update equation for e W, l 1., N1., p + N +1 is expressed as
Figure BDA0002374066980000063
η is the learning rate, which is a small constant.
Figure BDA0002374066980000064
The derivation is, i.e. the gradient is determined.
Step 4, respectively calculating the partial derivatives of E (k) and the imaginary part of the weight coefficient to obtain the gradient, i.e.
Figure BDA0002374066980000065
Step 5, performing gradient calculation to obtain partial derivatives of the real part of the complex weight:
Figure BDA0002374066980000066
obtaining partial derivatives of imaginary parts of complex weights:
Figure BDA0002374066980000071
step 6, mixing
Figure BDA0002374066980000072
And
Figure BDA0002374066980000073
expressed as sensitivity, respectively
Figure BDA0002374066980000074
The sensitivities in step 7 and step 6 should satisfy the relationship of Cauchy-Riemann equation:
Figure BDA0002374066980000075
and calculating to obtain the relation:
Figure BDA0002374066980000076
step 8, substituting the step 5, the step 6 and the step 7 into the step 4 to obtain
Figure BDA0002374066980000077
Is the conjugate of the complex number.
And 9, obtaining an update equation of the sensitive function by calculation as follows:
Figure BDA0002374066980000078
δlnis a kronecker function.
And step 10, replacing the steps 8 and 9 with the step 3 to obtain a weight updating equation:
Figure BDA0002374066980000079
and after the iteration of the algorithm is finished, updated weight is obtained so as to obtain optimized model prediction output, and the difference value between the model prediction output and the reference signal is used as modeling effect evaluation.
Step E, utilizing an inversion method, inputting the power amplifier output signal as a model, modeling the power amplifier input signal as a reference signal to obtain an inverse model of the power amplifier, namely the power amplifier digital predistorter, and then entering the step F;
step F, inputting an input signal x (n), entering a digital predistorter to obtain an output sequence signal z (n), processing the output sequence signal z (n) by a power amplifier model to obtain an output sampling signal v (n), and entering step G;
step G, obtaining absolute error signal | e (n) | according to e (n) ═ x (n) — v (n), and determining the predistorter effect according to the magnitude of | e (n) | as shown in fig. 2.
The pre-distorted signal is input into the dual-band power amplifier, the output signal and the power amplifier input signal present a linear relation, and the nonlinear distortion characteristic of the power amplifier is greatly improved. After the signal is subjected to predistortion treatment, the phase change of the signal is opposite to the phase change rule before the signal is subjected to predistortion treatment, after the signal is amplified by a dual-band power amplifier, the phase difference of an input signal and an output signal is basically 0, and the memory effect of the power amplifier is proved to be improved.
Referring to the attached figure 3 of the specification, the memory depth of the model is set to be 4, the feedback neuron is set to be 5, the CRTRL algorithm is adopted in the training algorithm, the error between the predicted output of the FCRNN model and the actual output of the power amplifier is small, and the NMSE is-46.51 dB, so that the model can well fit the nonlinear characteristic of the power amplifier, and the modeling effect of the F-class power amplifier is achieved under the conditions of training data 8000 and prediction data 4000.
Referring to fig. 4 of the drawings, this embodiment shows the power amplifier output power spectral density before and after digital predistortion, as a preferred embodiment of the present invention. Before the pre-distortion treatment, the original output power of the power amplifier is-33.87 dB. The ACPR of the output signal of the power amplifier after digital predistortion is-51.53 Bc, and the correction effect of 17.66dB is achieved. It can thus be shown that the digital predistortion technique effectively improves the spectral regeneration of the power amplifier output signal when excited by a WCDMA signal of 10M bandwidth.
Those skilled in the art will appreciate that those matters not described in detail in the present specification are well known in the art.

Claims (4)

1. A power amplifier digital predistortion method of a complex value full-connection recurrent neural network model is characterized by comprising the following steps:
(1) sending signal data x (n) to a hardware communication system, and acquiring an output signal y (n) of the radio frequency power amplifier through a feedback channel of the hardware communication system;
(2) performing auto-correlation synchronization between y (n) and x (n), and performing synchronous alignment processing on the input and output signals;
(3) after normalization processing is carried out on x (n) and y (n), a complex value full-connection recurrent neural network model is used for carrying out preliminary power amplifier modeling;
(4) updating parameters of the established initial power amplifier model by using a recursive learning training algorithm to obtain a final power amplifier model;
(5) by utilizing an inversion method, a power amplifier output signal is used as a model input, a power amplifier input signal is used as a reference signal for modeling, and an inverse model of the power amplifier is obtained, namely the inverse model of the power amplifier is a digital predistorter of the power amplifier;
(6) an input signal x (n) enters a power amplifier digital predistorter to obtain an output sequence signal z (n), and the output sequence signal z (n) is processed by a power amplifier model to obtain an output sampling signal v (n);
(7) obtaining absolute error signals | e (n) | according to e (n) ═ x (n) — v (n), and judging the effect of the predistorter according to the size of | e (n) |.
2. The method for digitally pre-distorting a power amplifier of a complex fully-connected recurrent neural network model of claim 1, wherein: in the complex value full-connected recurrent neural network model, the input, the weight and the output are complex numbers, and the complex value full-connected recurrent neural network model comprises N neurons, p external inputs and N feedback connecting lines, has a two-layer structure and is respectively an external input feedback layer and an output processing layer; in the network, the complex value of each neuron at the k-th time is output by yl(k) Denotes that, l ═ 1,. N,the external input is represented by p delay terms of s (k), which is a (1 × p) vector, where s (k) is the input sample sequence, and the total input vector P (k) of the entire network is represented by the output vector yl(k) S (k) and an offset input (1+ j) in series, as:
Figure FDA0002374066970000011
in the formula (I), the compound is shown in the specification,
Figure FDA0002374066970000021
(·)Ttranspose of the representation vector, (.)rAnd (·)iRepresenting the real and imaginary parts of a complex value or complex vector; for the l-th neuron, its weight is a vector of (p + N +1) × 1 dimensions, the same as the input vector dimensions, expressed as:
Figure FDA0002374066970000022
the complex matrix of the entire network is expressed by the formula: w ═ W1,...,wN]；
The output of each neuron is defined as: y isl(k)＝Φ(netl(k) Φ represents the complex-valued nonlinear activation function of the neuron, netl(k) Is the input of the activation function at time k, i.e. the linear sum of all the inputs of a certain node after applying weights, and is expressed as:
Figure FDA0002374066970000023
yl(k)＝Φr(netl(k))+jΦi(netl(k))＝ul(k)+jvl(k)。
3. the power amplifier digital predistortion method of the complex value full-connected recurrent neural network model according to claim 2, characterized in that: the activation function is an initial transcendental function Φ (z) ═ tanh (z).
4. The power amplifier digital predistortion method of the complex value full-connected recurrent neural network model according to claim 2, characterized in that: in the step (4), a recursive learning training algorithm is used for parameter updating, and the method specifically comprises the following steps:
step 1, supposing that the output of each neuron in the output layer is yt(k) N, with an output error of et(k) Error is represented by real part et r(k) And imaginary part et i(k) Composition, expressed as: e.g. of the typet(k)＝d(k)-yt(k)＝et r(k)+jet i(k)，et r(k)＝dr(k)-ut(k),et i(k)＝di(k)-vt(k)，d(k)＝dr(k)+jdi(k) Represents a reference signal;
step 2, taking the difference value of the reference signal and the output signal of the neuron as a cost function and expressing the difference value as
Figure FDA0002374066970000024
Step 3, weighting each weight w in the networkl,nAn update equation for e W, l 1., N1., p + N +1 is expressed as
Figure FDA0002374066970000025
η is the learning rate at which the user can learn,
Figure FDA00023740669700000210
to calculate the gradient;
Figure FDA0002374066970000026
is the conjugation of the complex number;
Figure FDA0002374066970000027
Figure FDA0002374066970000028
are respectively as
Figure FDA0002374066970000029
And
Figure FDA0002374066970000031
a simplified representation of;
Figure FDA0002374066970000032
δlnis a kronecker function.
CN202010059771.3A 2020-01-19 2020-01-19 Power amplifier digital predistortion method of complex-valued fully-connected recurrent neural network model Active CN111245375B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010059771.3A CN111245375B (en) 2020-01-19 2020-01-19 Power amplifier digital predistortion method of complex-valued fully-connected recurrent neural network model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010059771.3A CN111245375B (en) 2020-01-19 2020-01-19 Power amplifier digital predistortion method of complex-valued fully-connected recurrent neural network model

Publications (2)

Publication Number Publication Date
CN111245375A true CN111245375A (en) 2020-06-05
CN111245375B CN111245375B (en) 2023-06-06

Family

ID=70876285

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010059771.3A Active CN111245375B (en) 2020-01-19 2020-01-19 Power amplifier digital predistortion method of complex-valued fully-connected recurrent neural network model

Country Status (1)

Country Link
CN (1) CN111245375B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111988250A (en) * 2020-07-14 2020-11-24 清华大学 Digital predistortion structure, analog full-connection hybrid beam forming system and transmitter
CN112597705A (en) * 2020-12-28 2021-04-02 哈尔滨工业大学 Multi-feature health factor fusion method based on SCVNN
CN113468842A (en) * 2021-08-16 2021-10-01 重庆大学 Wideband digital predistortion algorithm based on vector quantization
CN113612455A (en) * 2021-08-16 2021-11-05 重庆大学 Digital predistortion system working method based on iterative learning control and principal curve analysis
WO2022166534A1 (en) * 2021-02-07 2022-08-11 大唐移动通信设备有限公司 Pre-distortion processing method and apparatus
CN115378446A (en) * 2022-10-25 2022-11-22 北京力通通信有限公司 Broadband digital predistortion system and method based on neural network

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070075770A1 (en) * 2002-04-23 2007-04-05 Maolin Long Base-band digital pre-distortion-based method for improving efficiency of rf power amplifier
CN101072220A (en) * 2006-05-08 2007-11-14 中兴通讯股份有限公司 Radical basis function neural network predistortion method for adaptive power amplifier
CN101320960A (en) * 2008-07-18 2008-12-10 东南大学 Power amplifier predistortion method of Hammerstein model based on fuzzy neural network
CN101997492A (en) * 2010-09-29 2011-03-30 东南大学 Simplified fuzzy neural network reinforced Wiener model based power amplifier predistortion method
CN102427438A (en) * 2011-11-29 2012-04-25 电子科技大学 Parameter training method for adaptive digital pre-distortion
EP2538553A1 (en) * 2011-06-21 2012-12-26 Alcatel Lucent Apparatus and method for mitigating impairments of a transmit signal
CN103731105A (en) * 2014-01-03 2014-04-16 东南大学 Amplifier digital pre-distortion device and method based on dynamic fuzzy neural network
WO2017118202A1 (en) * 2016-01-04 2017-07-13 中兴通讯股份有限公司 Software-and-hardware-coordinated digital pre-distortion method and device
CN110414565A (en) * 2019-05-06 2019-11-05 北京邮电大学 A kind of neural network method of cutting out based on Group Lasso for power amplifier
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

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070075770A1 (en) * 2002-04-23 2007-04-05 Maolin Long Base-band digital pre-distortion-based method for improving efficiency of rf power amplifier
CN101072220A (en) * 2006-05-08 2007-11-14 中兴通讯股份有限公司 Radical basis function neural network predistortion method for adaptive power amplifier
CN101320960A (en) * 2008-07-18 2008-12-10 东南大学 Power amplifier predistortion method of Hammerstein model based on fuzzy neural network
CN101997492A (en) * 2010-09-29 2011-03-30 东南大学 Simplified fuzzy neural network reinforced Wiener model based power amplifier predistortion method
EP2538553A1 (en) * 2011-06-21 2012-12-26 Alcatel Lucent Apparatus and method for mitigating impairments of a transmit signal
CN102427438A (en) * 2011-11-29 2012-04-25 电子科技大学 Parameter training method for adaptive digital pre-distortion
CN103731105A (en) * 2014-01-03 2014-04-16 东南大学 Amplifier digital pre-distortion device and method based on dynamic fuzzy neural network
WO2017118202A1 (en) * 2016-01-04 2017-07-13 中兴通讯股份有限公司 Software-and-hardware-coordinated digital pre-distortion method and device
CN110414565A (en) * 2019-05-06 2019-11-05 北京邮电大学 A kind of neural network method of cutting out based on Group Lasso for power amplifier
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 (2)

* Cited by examiner, † Cited by third party
Title
刘月等: "简化的滤波器查找表与神经网络联合预失真方法", 《计算机工程》 *
邓洪敏等: "基于BP神经网络的功放自适应预失真", 《通信学报》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111988250A (en) * 2020-07-14 2020-11-24 清华大学 Digital predistortion structure, analog full-connection hybrid beam forming system and transmitter
CN111988250B (en) * 2020-07-14 2023-03-10 清华大学 Simulation full-connection hybrid beam forming system and transmitter
CN112597705A (en) * 2020-12-28 2021-04-02 哈尔滨工业大学 Multi-feature health factor fusion method based on SCVNN
CN112597705B (en) * 2020-12-28 2022-05-24 哈尔滨工业大学 Multi-feature health factor fusion method based on SCVNN
WO2022166534A1 (en) * 2021-02-07 2022-08-11 大唐移动通信设备有限公司 Pre-distortion processing method and apparatus
CN113468842A (en) * 2021-08-16 2021-10-01 重庆大学 Wideband digital predistortion algorithm based on vector quantization
CN113612455A (en) * 2021-08-16 2021-11-05 重庆大学 Digital predistortion system working method based on iterative learning control and principal curve analysis
CN113612455B (en) * 2021-08-16 2024-01-26 重庆大学 Working method of digital predistortion system based on iterative learning control and main curve analysis
CN115378446A (en) * 2022-10-25 2022-11-22 北京力通通信有限公司 Broadband digital predistortion system and method based on neural network
CN115378446B (en) * 2022-10-25 2023-01-10 北京力通通信有限公司 Broadband digital predistortion system and method based on neural network

Also Published As

Publication number Publication date
CN111245375B (en) 2023-06-06

Similar Documents

Publication Publication Date Title
CN111245375B (en) Power amplifier digital predistortion method of complex-valued fully-connected recurrent neural network model
Hu et al. Convolutional neural network for behavioral modeling and predistortion of wideband power amplifiers
CN103299542B (en) For the orthogonal basis function collection of digital predistorter
CN102487367B (en) Adaptive amplifying digital baseband pre-distortion method
CN100594669C (en) Power amplifier predistortion method of Hammerstein model based on fuzzy neural network
WO2015096735A1 (en) Digital pre-distortion parameter obtaining method and pre-distortion system
CN111490737B (en) Nonlinear compensation method and equipment for power amplifier
CN110765720A (en) Power amplifier predistortion method of complex value assembly line recurrent neural network model
CN106982183A (en) A kind of adaptive blind equalization method in high-speed digital transmission system
CN113676426B (en) Intelligent digital predistortion system and method for dynamic transmission
CN105471784A (en) Digital predistortion method of jointly compensating for IQ imbalance and PA non-linearity
EP2641324A1 (en) Joint process estimator with variable tap delay line for use in power amplifier digital predistortion
CN102075469B (en) Estimation method for signal delay time of digital pre-distortion system
CN111859795A (en) Polynomial-assisted neural network behavior modeling system and method for power amplifier
CN101072220A (en) Radical basis function neural network predistortion method for adaptive power amplifier
CN115278746B (en) Self-adaptive fast error correction digital predistortion method for 5G broadband power amplifier
Jiang et al. Block-oriented time-delay neural network behavioral model for digital predistortion of RF power amplifiers
Tripathi et al. Swish activation based deep neural network predistorter for RF-PA
CN111200470B (en) High-order modulation signal transmission control method suitable for being interfered by nonlinearity
CN105471783B (en) Mimo system transmitting terminal digital pre-distortion optimization method based on list entries
CN110086438B (en) Digital predistortion system and method for passive multi-beam transmitter
CN111884602B (en) Power amplifier predistortion method based on single-output-node neural network
CN110533169A (en) A kind of digital pre-distortion method and system based on complex value neural network model
CN115913140B (en) Piecewise polynomial digital predistortion device and method for controlling operation precision
Qiu et al. A novel neural network based equalizer for nonlinear power amplifiers

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