CN106849882B - A kind of ultra-wideband low-noise amplifier based on artificial neural network - Google Patents

A kind of ultra-wideband low-noise amplifier based on artificial neural network Download PDF

Info

Publication number
CN106849882B
CN106849882B CN201710049060.6A CN201710049060A CN106849882B CN 106849882 B CN106849882 B CN 106849882B CN 201710049060 A CN201710049060 A CN 201710049060A CN 106849882 B CN106849882 B CN 106849882B
Authority
CN
China
Prior art keywords
data
neural network
artificial neural
control voltage
ultra
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
Application number
CN201710049060.6A
Other languages
Chinese (zh)
Other versions
CN106849882A (en
Inventor
刘洋
顾珣
莫雁杰
高宝玲
雷旭
于奇
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.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
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 University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201710049060.6A priority Critical patent/CN106849882B/en
Publication of CN106849882A publication Critical patent/CN106849882A/en
Application granted granted Critical
Publication of CN106849882B publication Critical patent/CN106849882B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • H03F1/26—Modifications of amplifiers to reduce influence of noise generated by amplifying elements
    • 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
    • H03F1/42—Modifications of amplifiers to extend the bandwidth
    • 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
    • H03F3/21—Power amplifiers, e.g. Class B amplifiers, Class C amplifiers with semiconductor devices only
    • 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
    • H03F3/24—Power amplifiers, e.g. Class B amplifiers, Class C amplifiers of transmitter output stages
    • H03F3/245—Power amplifiers, e.g. Class B amplifiers, Class C amplifiers of transmitter output stages with semiconductor devices only
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03F—AMPLIFIERS
    • H03F2200/00—Indexing scheme relating to amplifiers
    • H03F2200/294—Indexing scheme relating to amplifiers the amplifier being a low noise amplifier [LNA]
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03F—AMPLIFIERS
    • H03F2200/00—Indexing scheme relating to amplifiers
    • H03F2200/372—Noise reduction and elimination in amplifier
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03F—AMPLIFIERS
    • H03F2200/00—Indexing scheme relating to amplifiers
    • H03F2200/451—Indexing scheme relating to amplifiers the amplifier being a radio frequency amplifier

Landscapes

  • Engineering & Computer Science (AREA)
  • Power Engineering (AREA)
  • Amplifiers (AREA)

Abstract

A kind of ultra-wideband low-noise amplifier based on artificial neural network, belongs to field of radio frequency integrated circuits.Including control voltage generating module and ultra-wideband low-noise amplifier module, ultra-wideband low-noise amplifier module includes the variable resistance being made of the second NMOS tube M3 and the second PMOS tube M4, voltage generating module is controlled by sensor, data selector, input data processing unit, artificial neural network and output data processing unit composition, utilize artificial neural network itself unintentional nonlinearity mapping ability, enable network Nonlinear Function Approximation with high precision, to generate to frequency in the control voltage of non-linear relation and be added to the grid of the second NMOS tube M3 and the second PMOS tube M4 and change its variable resistance resistance value, to guarantee bandwidth, maximumlly improve the noise data NF of circuit under conditions of gain and good fitting.

Description

A kind of ultra-wideband low-noise amplifier based on artificial neural network
Technical field
The invention belongs to field of radio frequency integrated circuits, and in particular to a kind of superwide band low noise based on artificial neural network Amplifier.
Background technique
With the development of short-distance wireless communication, requirement of the people to high-speed radiocommunication system performance is higher and higher, more Large capacity, faster speed and safer communication are the inevitable development trend of short-distance wireless communication.And wirelessly communicate system It unites and be unable to do without receiver, the radio-frequency module of front end is low-noise amplifier LNA in receiver, it is to entire receiver and entirely Communication system has important influence.In order to meet the needs of high-speed transfer, it is desirable that low-noise amplifier LNA can be in wide frequency band Work in range, so having caused the research to ultra-wideband low-noise amplifier.
In addition to this, low-noise amplifier LNA is as the biggish module of power consumption in entire receiver, in order to guarantee its continuation of the journey Ability needs that it is made to have low power consumption.For UWB (ultra wideband) system, that is, radio ultra wide band system, low-power consumption is Its basic demand.But due to there is very big noise in input signal, according to system noise cascading equations it is found that in receiving end Low-noise amplifier must provide enough gains to guarantee that rear class noise will not cause excessive influence to system performance, together When, enough gains need to consume high power consumption and are just able to achieve.Therefore low-noise amplifier LNA gain requirement and power consumption requirements it Between there are the contradiction of certain relationship, how to increase as far as possible while reducing power consumption gain be applied to it is low in UWB system The important topic of noise amplifier LNA design.
The realization of ultra wide band low noise amplifier circuit usually has 3 kinds of modes: the first is using such as article " A The distribution that Monolithic DC-70-GHz Broadband Distributed Amplifier Using 90-nm " is reported Formula structure expands bandwidth；Second is using such as article " Bandwidth extension techniques for CMOS The inductor peaking technology that amplifiers " is reported；The third is to utilize article " Wide-Band CMOS Low-Noise The resistive degeneration noise cancellation technique that Amplifier Exploiting Thermal Noise Canceling " is reported. However, the first implementation can occupy very big chip area, and power consumption is big, noiseproof feature is poor；The second way is not easy Realize broadband input matching；The third mode is difficult to realize high flat gain.To solve this problem, patent The ultra wide band low-power consumption noise amplifier that a kind of automatic biasing is referred in CN10479919A, realizes in 0.2-6GHz frequency model It encloses the LNA of interior work and keeps 16 ± 1.3dB gain, the noise coefficient of < 2.8dB and good input matching.However it is traditional Circuit design has the limitation of itself technique and design always, how to further increase the bandwidth of LNA, reduces noise, improves increasing Benefit reduces power consumption and maintains good gain flatness, is the key point of ultra-wideband low-noise amplifier design.
Summary of the invention
For existing ultra-wideband low-noise amplifier because traditional handicraft and design limitation can not provide work where it The shortcomings that making the optimum noise data NF of frequency range, the present invention provides a kind of superwide band low noises based on artificial neural network to put Big device.
Technical scheme is as follows:
A kind of ultra-wideband low-noise amplifier based on artificial neural network, including control voltage generating module and ultra wide band Low noise amplifier module,
The ultra-wideband low-noise amplifier module includes low-frequency gain amplifying circuit and high-frequency gain amplifying circuit, described The input terminal of high-frequency gain amplifying circuit connects the output end of the low-frequency gain amplifying circuit, and output end is as the ultra-wide Output end with low noise amplifier module, the low-frequency gain amplifying circuit include by the second NMOS tube M3 and the second PMOS tube The variable resistance that M4 is constituted；
The control voltage generating module includes sensor, data selector, input data processing unit, artificial neural network Network and output data processing unit,
The input terminal of sensor connects the output end of the ultra-wideband low-noise amplifier module, and it is low to acquire the ultra wide band Frequency data f, noise data NF, matched data S11 and the gain data S21 of noise amplifier module；
The output end of data selector, first input end and sensor connects, the second input terminal connection control voltage letter Number Contr, data selector are selected to obtain under the control of control voltage signal Contr to the collected data of sensor Pre-input data；
Input data processing unit, first input end connect the output end of data selector, the connection control of the second input terminal Voltage signal Contr processed, output end are connected with the first input end of artificial neural network, and pre-input data handle The input data X that can be directly handled to artificial neural network；
Artificial neural network, the second input terminal connection control voltage signal Contr, output end connect at output data The input terminal for managing unit, to input data X it is anticipated that the nonlinear function of setting under the control of control voltage signal Contr Nonlinear response is made in transformation V=g (X), generates the digital signal of non-linear control voltage；
Output data processing unit, input terminal link the output end of artificial neural network, artificial neural network are exported The digital signal of non-linear control voltage be converted to voltage signal, i.e. control voltage V, it is low that output end connects the ultra wide band The grid of the second NMOS tube M3 and the second PMOS tube M4 adjust the resistance value of the variable resistance of its composition in noise amplifier module, together When feed back to input data processing unit.
Specifically, the control voltage signal Contr is state switching signal, control artificial neural network, data selection Device and input data processing unit switch between learning state and working condition.
Specifically, being input to defeated when control voltage signal Contr control input data processing unit is in learning state The data for entering data processing unit include frequency data f, control voltage V, noise data NF, matched data S11 and gain data S21, then input data X=(f, V, NF, S11, S21)；When control voltage signal Contr control data selector is in work shape When state, the data for being input to data selector only have frequency data f, then input data X=(f).
Specifically, the low-frequency gain amplifying circuit includes the first NMOS tube M1, the first PMOS tube M2, the second NMOS tube M3, the second PMOS tube M4 and the first inductance L1,
The one of the grid of first PMOS tube M2, the drain electrode of the second NMOS tube M3 and the second PMOS tube M4 and the first inductance L1 End, which is connected, constitutes the input terminal of the ultra-wideband low-noise amplifier module, and the other end of the first inductance L1 connects the first NMOS tube The grid of M1,
The grid of second NMOS tube M3 and the second PMOS tube M4 connect the output end of the output data processing unit, and second The source electrode of NMOS tube M3 and the second PMOS tube M4 are connect and with the drain electrode of the first NMOS tube M1 and the first PMOS tube M2 as described The output end of low-frequency gain amplifying circuit,
The source electrode of first NMOS tube M1 is grounded GND, and the source electrode of the first PMOS tube M2 meets supply voltage VDD.
Specifically, the high-frequency gain amplifying circuit includes the second inductance L2, third inductance L3, third NMOS tube M5, the Three PMOS tube M6,
The output end of a termination low-frequency gain amplifying circuit of second inductance L2, the other end connect third NMOS tube The grid of M5 and third PMOS tube M6,
One end of the drain electrode connection third inductance L3 of third PMOS tube M6 is as the ultra-wideband low-noise amplifier module Output end and connect with the input terminal of sensor of the control voltage generating module, source electrode meets supply voltage VDD,
The drain electrode of third NMOS tube M5 connects the other end of third inductance L3, and source electrode is grounded GND.
A kind of course of work of ultra wide band low noise amplifier based on artificial neural network proposed by the present invention is divided into three Stage, specifically:
First stage: collecting training data
At each different frequency fi, i is temperature label, is constituted by the second PMOS tube M4 and the second NMOS tube M3 Variable resistance both ends apply the control right ± Vi of voltage, the grid of the second NMOS tube M3 connects positive controling voltage, the second PMOS tube The grid of M4 connects negative control voltage, so that circuit meets matching and gain requirement under this frequency in different Frequency point fi When reach minimum noise data NF, record data (Vi, fi), obtain training sample Y=(Vi, fi)；Training sample Y is at this time Two-dimensional array, wherein input of the f as artificial neural network, V are exported as the target of artificial neural network.
Second stage: the training (study) of artificial neural network
The stage, control voltage signal Contr control artificial neural network is in learning state, while control passes through data The data of selector are the noise data NF obtained in the collecting training data stage.The training sample data Y of first stage acquisition =(Vi, fi) is input to artificial neural network after input data processing unit processes, and artificial neural network is to input variable X= (f, V, NF, S11, S21) is responded, and generates network output, is then compared to network output and target output V, is worked as the two Error when being unsatisfactory for preset required precision, artificial neural network adjusts network weight W, until error is less than preset essence Degree, training terminate.
Phase III: work
The stage, control voltage signal Contr control artificial neural network are in running order.Control voltage signal The data that Contr control is input to input data processing unit are X=(f), and artificial neural network is according in second stage training Good network weight W is to input data X=(f) it is anticipated that the non-linear function transformation V=g (X) of setting makes non-linear sound Answer, generate non-linear control voltage V=g (f), the control voltage controlled after output data processing unit the second PMOS tube M4, The grid end voltage of second NMOS tube M3, so that the variable resistance resistance value being made of the second PMOS tube M4 and the second NMOS tube M3 be made to become Change so that circuit obtains meeting the optimal noise data NF under gain, matching condition.
The invention has the benefit that artificial neural network is used for low-noise amplifier by the present invention, it is super to overcome tradition Wideband low noise amplifier was designed due to feedback resistance immutable the shortcomings that can not further being promoted to circuit performance, was passed through Control voltage generating module in artificial neural network itself unintentional nonlinearity mapping ability enable the network with high precision Nonlinear Function Approximation, to generate the control voltage to frequency in nonlinear function, the voltage is brilliant by the 2nd PMOS of control The grid end voltage of body pipe M4 and the second NMOS transistor M3, to make the second PMOS transistor M4 and the second NMOS transistor M3 group At variable resistance change in resistance so that circuit is met the optimal noise under gain, matching condition at respective frequencies Coefficient NF；Second level high-frequency gain amplifying circuit is effectively opened up using the inductor peaking effect of the second inductance L2 and third inductance L3 This circuit high-frequency gain bandwidth again is opened up.
Detailed description of the invention
Fig. 1 is the structural representation of the ultra-wideband low-noise amplifier proposed by the present invention for having been based on artificial neural network Figure.
Fig. 2 is the training schematic diagram of artificial neural network.
Fig. 3 is ultra wide band low noise in the ultra-wideband low-noise amplifier proposed by the present invention for having been based on artificial neural network The topology diagram of acoustic amplifier module.
Fig. 4 is the small signal simplification figure of ultra-wideband low-noise amplifier modular circuit.
Fig. 5 is a kind of gain of ultra-wideband low-noise amplifier based on artificial neural network proposed by the present invention, matching With noise factor simulation curve.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawing and apart from embodiment.
Core of the invention is that employment artificial neural networks are generated by the second PMOS transistor M4 and the second NMOS transistor Control voltage needed for the variable resistance both ends that M3 is constituted.Artificial neural network itself unintentional nonlinearity mapping ability makes this Network can Nonlinear Function Approximation with high precision, thus generate be in frequency nonlinear function control voltage, the voltage is logical Control the second PMOS transistor M4, the second NMOS transistor M3 gate voltage are crossed, so that variable resistance change in resistance be made to make circuit Obtain meeting the optimal noise coefficient NF under gain, matching condition.
It is as shown in Figure 1 the structure of the ultra-wideband low-noise amplifier proposed by the present invention for having been based on artificial neural network Schematic diagram, including control voltage generating module and ultra-wideband low-noise amplifier module.Wherein, control voltage generating module is by passing Sensor, data selector, input data processing unit, artificial neural network and output data processing unit composition, effect are Generate the control voltage to frequency in non-linear relation；The control voltage is low by output data processing unit output control ultra wide band The grid end voltage of second PMOS tube M4 and the second NMOS tube M3 in noise amplifier module, to make by the second PMOS tube M4 and The variable resistance change in resistance that two NMOS tube M3 are constituted makes circuit obtain meeting the optimal NF under gain, matching condition.
In the control voltage generating module, sensor is connected with input data processing unit, super for detecting and transmitting Frequency data f, noise data NF, matched data S11 and the gain data S21 of wideband low noise amplifier module.
The input terminal of input data processing unit is also connected with the control voltage in ultra-wideband low-noise amplifier module, defeated It is connected out with the input of artificial neural network, effect is to receive input data, and artificial neural network is processed into it can be with The data directly handled.
The output of artificial neural network terminates output data processing unit, and effect is in control voltage signal Contr control Under to input data x it is anticipated that the non-linear function transformation V=g (x) of setting makes nonlinear response, generate nonlinear Control Voltage；
The output of output data processing unit meets the second NMOS tube M3 and second in ultra-wideband low-noise amplifier module The control voltage of PMOS tube M4, effect are followed by the processing of the output data of artificial neural network to the second NMOS tube M3 and the The grid of two PMOS tube M4 is realized pair by control the second NMOS tube M3 and the second PMOS tube M4 in the resistance in linear work area The adjusting of circuit noise.
Wherein control voltage signal Contr is artificial neural network and data selector and input data processing unit State switching signal, the signal control artificial neural network and data selector and input data processing unit in learning state Switch between working condition.
When control voltage signal Contr control artificial neural network is in learning state, control voltage signal Contr control The data that system is input to input data processing unit are 5 dimensions, including frequency f, control voltage v, noise NF, gain S21, matching S11, both X=(V, f, NF, S21, S11)；And when control voltage signal Contr control artificial neural network is in running order When, the data that control voltage signal Contr control is input to input data processing unit only have frequency data f, it may be assumed that X=(f).
The course of work of the ultra wide band low noise amplifier based on artificial neural network is divided into three phases, specifically:
First stage: collecting training data
At each different frequency fi, i is temperature label, is constituted by the second PMOS tube M4 and the second NMOS tube M3 Variable resistance both ends apply the control right ± Vi of voltage, the grid of the second NMOS tube M3 connects positive controling voltage, the second PMOS tube The grid of M4 connects negative control voltage, so that circuit meets matching and gain requirement under this frequency in different Frequency point fi When reach minimum noise data NF, record data (Vi, fi), obtain training sample Y=(Vi, fi)；Training sample Y is at this time Two-dimensional array, wherein input of the f as artificial neural network, V are exported as the target of artificial neural network.
Second stage: the training (study) of artificial neural network
The stage, control voltage signal Contr control artificial neural network is in learning state, while control passes through data The data of selector are the noise data NF obtained in the collecting training data stage.The training sample data Y of first stage acquisition =(Vi, fi) is input to artificial neural network after input data processing unit processes, and artificial neural network is to input variable X= (f, V, NF, S11, S21) is responded, and generates network output, is then compared to network output and target output V, is worked as the two Error when being unsatisfactory for preset required precision, artificial neural network adjusts network weight W, until error is less than preset essence Degree, training terminate.
Phase III: work
The stage, control voltage signal Contr control artificial neural network are in running order.Control voltage signal The data that Contr control is input to input data processing unit are X=(f), and artificial neural network is according in second stage training Good network weight W makes nonlinear response to input data X=(f), generates the number letter of non-linear control voltage V=g (f) Number, the digital signal of the control voltage is converted to voltage signal after output data processing unit and controls the second PMOS tube M4, the The grid end voltage of two NMOS tube M3, to make the variable resistance change in resistance being made of the second PMOS tube M4 and the second NMOS tube M3 So that circuit obtains meeting the optimal noise data NF under gain, matching condition.
Ultra-wideband low-noise amplifier module ultra-wideband low-noise amplifier modular circuit of the invention as shown in figure 3, by The low-frequency gain amplifying circuit of the first order and the high-frequency gain amplifying circuit composition of the second level；
Wherein, low-frequency gain amplifying circuit includes the first NMOS tube M1, the first PMOS tube M2, the second NMOS tube M3, second PMOS tube M4 and the first inductance L1, the grid of the first PMOS tube M2, the second NMOS tube M3 and the second PMOS tube M4 drain electrode and One end of first inductance L1, which is connected, constitutes the input terminal of the ultra-wideband low-noise amplifier module, the other end of the first inductance L1 The grid of the first NMOS tube M1 is connected, the grid connection of the second NMOS tube M3 and the second PMOS tube M4 the output data processing are single The drain electrode of the source electrode and the first NMOS tube M1 and the first PMOS tube M2 of the output end of member, the second NMOS tube M3 and the second PMOS tube M4 It connects and the output end as the low-frequency gain amplifying circuit, the source electrode of the first NMOS tube M1 is grounded GND, the first PMOS tube M2 Source electrode meet supply voltage VDD.
This structural reference article A DC-11.5GHz Low-Power, Wideband of Shih-Fong Chao Amplifier Using Splitting-Load Inductive Peaking Technique, and carried out in its structure Optimization, automatic biasing, which is utilized, in the ultra wide band low noise amplifier circuit that the present invention designs has avoided the design of biasing circuit and has simplified Circuit design and constrain the power consumption of circuit.
Wherein, high-frequency gain amplifying circuit includes the second inductance L2, third inductance L3, third NMOS tube M5, the 3rd PMOS The output end of a termination low-frequency gain amplifying circuit of pipe M6, the second inductance L2, the other end connect third NMOS tube M5 With the grid of third PMOS tube M6, one end of the drain electrode connection third inductance L3 of third PMOS tube M6 is as the ultra wide band low noise The output end of acoustic amplifier module is simultaneously connect with the input terminal of the sensor of the control voltage generating module, and source electrode connects power supply Voltage VDD, the drain electrode of third NMOS tube M5 connect the other end of third inductance L3, and source electrode is grounded GND.
Second level high-frequency gain amplifying circuit is effective using the inductor peaking effect of the second inductance L2 and third inductance L3 This circuit high-frequency gain bandwidth again is expanded.
By the small-signal model of the available circuit of Fig. 3, as shown in figure 4, Cgs1, Cgs2, Cgs5, Cgs6 are respectively in figure The first NMOS tube of NMOS M1, the first PMOS tube M2, third NMOS tube M5, third PMOS tube M6 grid source between parasitic capacitance, Cgd1, Cgd2, Cgd3, Cgd4 are the first NMOS tube M1, the first PMOS tube M2, third NMOS tube M5, third PMOS tube M6 respectively Parasitic capacitance between grid leak, R ˊ are the variable resistances being made of the second PMOS tube M4, the second NMOS tube M3, ro1, ro2, ro5, Ro6 is the output resistance of the first NMOS tube M1, the first PMOS tube M2, third NMOS tube M5, third PMOS tube M6 respectively, gm1, Gm2, gm3, gm4 are the small signal of the first NMOS tube M1, the first PMOS tube M2, third NMOS tube M5, third PMOS tube M6 respectively Mutual conductance.
By small signal circuit figure, we can be derived by the input resistance R of circuitin
And first order circuit gain
Wherein s=jw=j2 π f
For secondary structure, the overall noise factor of circuit
Wherein NF1It is the noise coefficient of first order circuit, NF2It is the noise coefficient of second level circuit, AP1It is first order electricity The available power gain on road.It can be seen that the noise coefficient of the first order plays leading role in the overall noise factor of circuit.The In primary circuit, R ˊ plays an important role in extension low-noise amplifier bandwidth as feedback resistance, however resistance feedback Network can generate the thermal noise of itself, therefore pass through the noise coefficient of adjusting R ˊ size also adjustable circuit.
Control voltage needed for generating variable resistance both ends by artificial neural network, artificial neural network itself are intrinsic Non-linear mapping capability enables network Nonlinear Function Approximation with high precision, so that generating to frequency is in nonlinear function Control voltage, the voltage by control the second PMOS transistor M4, the second NMOS transistor M3 gate voltage, to make second The variable resistance change in resistance of PMOS transistor M4, the second NMOS transistor M3 composition makes circuit obtain meeting gain, matching Under the conditions of optimal noise coefficient NF.
Fig. 5 illustrates the ultra-wideband low-noise amplifier based on artificial neural network for the optimum results of circuit.
It can be seen that the present invention overcomes legacy ultra-wideband low-noise amplifier design due to feedback resistance is immutable can not The shortcomings that further being promoted to circuit performance can achieve at respective frequencies most by the control of artificial neural network Low noise makes circuit performance be greatly improved.
Those skilled in the art disclosed the technical disclosures can make various do not depart from originally according to the present invention Various other specific variations and combinations of essence are invented, these variations and combinations are still within the scope of the present invention.

Claims (5)

1. a kind of ultra-wideband low-noise amplifier based on artificial neural network, including control voltage generating module and ultra wide band it is low Noise amplifier module, which is characterized in that
The ultra-wideband low-noise amplifier module includes low-frequency gain amplifying circuit and high-frequency gain amplifying circuit, the high frequency The input terminal of gain amplifying circuit connects the output end of the low-frequency gain amplifying circuit, and output end is low as the ultra wide band The output end of noise amplifier module, the low-frequency gain amplifying circuit include by the second NMOS tube (M3) and the second PMOS tube (M4) variable resistance constituted；
The control voltage generating module include sensor, data selector, input data processing unit, artificial neural network and Output data processing unit,
The input terminal of sensor connects the output end of the ultra-wideband low-noise amplifier module, acquires the superwide band low noise Frequency data f, noise data NF, matched data S11 and the gain data S21 of amplifier module；
The output end of data selector, first input end and sensor connects, the second input terminal connection control voltage signal Contr, data selector are selected to obtain pre- under the control of control voltage signal Contr to the collected data of sensor Input data；
Input data processing unit, first input end connect the output end of data selector, the second input terminal connection control electricity Signal Contr is pressed, output end is connected with the first input end of artificial neural network, in the control of control voltage signal Contr It is lower that pre-input data are handled to obtain the input data X that artificial neural network can be handled directly；
Artificial neural network, the second input terminal connection control voltage signal Contr, it is single that output end connects output data processing The input terminal of member, to input data X it is anticipated that the non-linear function transformation of setting under the control of control voltage signal Contr V=g (X) makes nonlinear response, generates the digital signal of non-linear control voltage；
Output data processing unit, input terminal connect the output end of artificial neural network, by the non-of artificial neural network output The digital signal of Linear Control voltage is converted to voltage signal, i.e., control voltage V, output end connect the superwide band low noise The grid of the second NMOS tube (M3) and the second PMOS tube (M4) adjusts the resistance value of the variable resistance of its composition in amplifier module, together When feed back to input data processing unit.
2. a kind of ultra-wideband low-noise amplifier based on artificial neural network according to claim 1, which is characterized in that The control voltage signal Contr is state switching signal, control artificial neural network, data selector and input data processing Unit switches between learning state and working condition.
3. a kind of ultra-wideband low-noise amplifier based on artificial neural network according to claim 2, which is characterized in that When control voltage signal Contr control input data processing unit is in learning state, it is input to input data processing unit Data include frequency data f, control voltage V, noise data NF, matched data S11 and gain data S21, then input data X =(f, V, NF, S11, S21)；When control voltage signal Contr control data selector is in running order, it is input to data The data of selector only have frequency data f, then input data X=(f).
4. a kind of ultra-wideband low-noise amplifier based on artificial neural network according to claim 1, which is characterized in that The low-frequency gain amplifying circuit includes the first NMOS tube (M1), the first PMOS tube (M2), the second NMOS tube (M3), the 2nd PMOS (M4) and the first inductance (L1) are managed,
The grid of first PMOS tube (M2), the drain electrode of the second NMOS tube (M3) and the second PMOS tube (M4) and the first inductance (L1) One end be connected and constitute the input terminal of the ultra-wideband low-noise amplifier module, the other end connection first of the first inductance (L1) The grid of NMOS tube (M1),
Second NMOS tube (M3) connects the output end of the output data processing unit with the grid of the second PMOS tube (M4), and second The source electrode of NMOS tube (M3) and the second PMOS tube (M4) is connect simultaneously with the drain electrode of the first NMOS tube (M1) and the first PMOS tube (M2) As the output end of the low-frequency gain amplifying circuit,
The source electrode of first NMOS tube (M1) is grounded (GND), and the source electrode of the first PMOS tube (M2) meets supply voltage (VDD).
5. a kind of ultra-wideband low-noise amplifier based on artificial neural network according to claim 1 or 4, feature exist In the high-frequency gain amplifying circuit includes the second inductance (L2), third inductance (L3), third NMOS tube (M5), the 3rd PMOS It manages (M6),
The output end of the one termination low-frequency gain amplifying circuit of the second inductance (L2), the other end connect third NMOS tube (M5) and the grid of third PMOS tube (M6),
One end of drain electrode connection third inductance (L3) of third PMOS tube (M6) is as the ultra-wideband low-noise amplifier module Output end and connect with the input terminal of sensor of the control voltage generating module, source electrode meets supply voltage (VDD),
The drain electrode of third NMOS tube (M5) connects the other end of third inductance (L3), and source electrode is grounded (GND).
CN201710049060.6A 2017-01-23 2017-01-23 A kind of ultra-wideband low-noise amplifier based on artificial neural network Active CN106849882B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710049060.6A CN106849882B (en) 2017-01-23 2017-01-23 A kind of ultra-wideband low-noise amplifier based on artificial neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710049060.6A CN106849882B (en) 2017-01-23 2017-01-23 A kind of ultra-wideband low-noise amplifier based on artificial neural network

Publications (2)

Publication Number Publication Date
CN106849882A CN106849882A (en) 2017-06-13
CN106849882B true CN106849882B (en) 2019-11-01

Family

ID=59119735

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710049060.6A Active CN106849882B (en) 2017-01-23 2017-01-23 A kind of ultra-wideband low-noise amplifier based on artificial neural network

Country Status (1)

Country Link
CN (1) CN106849882B (en)

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5517154A (en) * 1995-01-13 1996-05-14 Tektronix, Inc. Split-path linear isolation circuit apparatus and method
CN1095614C (en) * 1998-05-25 2002-12-04 中国石油天然气总公司地球物理勘探局 Amplifier noise reducing method and ultralow-noise pre-amplifier
CN101465618B (en) * 2007-12-19 2010-12-08 中国科学院微电子研究所 Ultra-wideband low-noise amplifier with double input routes
CN104779919B (en) * 2015-05-04 2018-03-02 中国电子科技集团公司第五十四研究所 A kind of ultra wide band low-power consumption low-noise amplifier of automatic biasing
CN105141263A (en) * 2015-09-29 2015-12-09 株洲宏达天成微波有限公司 Multi-band low-noise amplification method and multi-band low-noise amplifier

Also Published As

Publication number Publication date
CN106849882A (en) 2017-06-13

Similar Documents

Publication Publication Date Title
CN101741316B (en) Gain variable broadband radio frequency low-noise amplifier
CN103051298B (en) Programmable Gain Amplifier Circuit and programmable gain amplifier
CN107592082B (en) Dual-mode dual-frequency secondary current multiplexing low-noise amplifier
CN101316095B (en) Gain-controllable wide-band low-noise amplifier
CN102790595B (en) Single ended differential gain amplifier with configurable radio frequency broadband gain
TWI323555B (en) Low noise amplifier and related method
WO2012048544A1 (en) Amplifier and implementation method thereof
CN103051299B (en) Programmable gain amplifier applicable to transmitting end of communication system
CN103532497B (en) A kind of ultra-wideband low-noise amplifier using inductance compensation technology
KR20090084349A (en) Reconfigurable low noise amplifier using capacitor dividing feedback
CN104270100B (en) A kind of low-power consumption low-noise amplifier for strengthening technology using positive feedback technique and active transconductance
CN101951229A (en) Multi-gain mode low-noise amplifier
CN103117711A (en) Monolithic integrated radio frequency high-gain low-noise amplifier
CN102916666A (en) Broadband programmable gain amplifier
CN101888210A (en) Variable gain low-noise amplifier
CN105281682A (en) Low-power-consumption bidirectional noise-reducing low-noise amplifier
CN102801389A (en) Ultra-low power consumption low-noise amplifier
CN109257022A (en) A kind of working frequency levels off to fT/ 2 broad band amplifier
CN102916667B (en) A kind of broadband programmable gain amplifier of 2dB step-length
CN111478671A (en) Novel low-noise amplifier applied to Sub-GHz frequency band
CN107819445A (en) A kind of big output voltage swing drive circuit of high speed
CN106385239A (en) Gain-adjustable CMOS (Complementary Metal-Oxide-Semiconductor Transistor) broadband low-noise amplifier
CN106849882B (en) A kind of ultra-wideband low-noise amplifier based on artificial neural network
CN103944518B (en) Wideband low-noise amplifier
CN106936399B (en) A kind of consumption high gain high linearity broadband low-noise amplifier

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