A closed-loop detection noise modeling and suppression method

Abstract

The application discloses a closed-loop detection noise modeling and suppression method, comprising the following steps: constructing a voltage sensor equivalent circuit model, an automatic gain control module and a phase-locked loop double closed-loop circuit basic model; building a feedthrough cancellation circuit model and connecting the feedthrough cancellation circuit model with the voltage sensor equivalent circuit model; filtering a voltage signal as an input of a first band-pass filter; performing phase control and amplitude modulation on the filtered voltage signal; feeding back the signal after the phase control and the amplitude modulation to a second band-pass filter for filtering; inputting an output of the second band-pass filter to a driving electrode of the voltage sensor equivalent circuit model and coupling the output to an output of the voltage sensor equivalent circuit model; and performing adaptive filtering by using a time series-based adaptive Kalman filter to obtain a voltage signal after noise reduction. The application constructs a complete noise suppression system, and combines a feedthrough capacitor active compensation technology and an intelligent data processing algorithm to realize a high-precision and low-noise voltage measurement function.

Description

Technical Field

[0001] This invention relates to the field of noise suppression technology, specifically to a method for closed-loop detection noise modeling and suppression. Background Technology

[0002] In the development of equivalent circuit models for large-range voltage sensors, closed-loop detection technology is the core means to improve measurement accuracy and stability.

[0003] Current technologies typically suffer from the following problems: Traditional voltage sensor equivalent circuit models (such as resistive voltage dividers and capacitive voltage dividers) are susceptible to inherent circuit noise (1 / f noise, thermal noise) and external interference (electromagnetic coupling, mechanical vibration) over a wide measurement range. A single closed-loop circuit struggles to balance dynamic range and phase stability, especially under high-speed transient signals, where gain oscillations or phase-locked loop (PLL) loss of synchronization are common. Switching power supply ripple and ground loop noise can couple to the signal chain through the PDN network, reducing the signal-to-noise ratio. Regarding crosstalk compensation, feedthrough capacitors between the drive and detection channels introduce crosstalk; traditional shielding methods can only suppress it by 40dB, failing to meet high-precision requirements. Existing solutions often rely on post-processing algorithms for noise suppression, neglecting optimization at the hardware layout and closed-loop control level. Power integrity design lacks dynamic adjustment capabilities, cannot adapt to sudden load changes in data processing, and lacks a collaborative optimization architecture. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, this invention provides a closed-loop detection noise modeling and suppression method that solves the problems of existing technologies lacking dynamic adjustment capabilities, being unable to adapt to data processing with sudden load changes, and failing to form a collaborative optimization architecture.

[0005] To achieve the aforementioned objectives, the present invention employs the following technical solution: a closed-loop detection noise modeling and suppression method, comprising: Construct the equivalent circuit model of the voltage sensor, the basic model of the automatic gain control module and the phase-locked loop dual closed-loop circuit; A feedthrough cancellation circuit model was built and connected to the equivalent circuit model of the voltage sensor to remove power supply noise; The electric field signal is detected by the sensing electrode of the equivalent circuit model of the voltage sensor and converted into a current signal. The current signal is converted into a voltage signal through the inductive electrode IV conversion module; The voltage signal is used as the input of the first bandpass filter for filtering to obtain the filtered voltage signal. Phase-locked loop (PLL) and automatic gain control (AGC) modules are used to perform phase control and amplitude modulation on the filtered voltage signal, respectively. The signal after phase control and amplitude modulation is fed back to the second bandpass filter for filtering to remove noise introduced by the phase-locked loop and automatic gain control module; The output of the second bandpass filter is input to the driving electrode of the equivalent circuit model of the voltage sensor and coupled to the output of the equivalent circuit model of the voltage sensor. The output of the coupled voltage sensor equivalent circuit model is adaptively filtered using a time-series-based adaptive Kalman filter to obtain a denoised voltage signal.

[0006] The beneficial effects of this invention are as follows: 1. By constructing an equivalent circuit model of the electric field sensor, a basic model of the automatic gain control and phase-locked loop dual closed-loop circuit, as well as a noise model and a propagation path loop model, a comprehensive and detailed model of the system was achieved. Using Cadence software simulation to obtain the noise of each module of the AGC and PLL and importing it into the system loop analysis for verification, the noise source and propagation path can be accurately located, laying a solid theoretical and data foundation for noise suppression.

[0007] 2. To address external co-frequency interference, a feedthrough cancellation circuit model was constructed based on feedthrough effect research and circuit equivalent models, avoiding erroneous output interference detection results when oscillation is not yet activated. A power supply circuit was built based on power integrity theory, effectively reducing voltage ripple noise interference to the operational amplifier and improving circuit stability. Simultaneously, the construction of a bandpass filter allows for targeted filtering of sensor output signals, reducing sideband noise and improving signal purity in multiple dimensions.

[0008] 3. The time-series-based adaptive Kalman filter algorithm, through ARMA time-series modeling (including data preprocessing, order determination, and parameter estimation) and adaptive Kalman filter iteration (state space construction, adaptive mechanism, and filter iteration), can adaptively filter out phase noise in the sensor output modulation signal. Compared with traditional fixed algorithms, it can dynamically adapt to signal changes, significantly improve the phase noise suppression accuracy and system adaptability, enhance the reliability and accuracy of detection results, and improve the performance and application value of the entire voltage sensor detection system. Attached Figure Description

[0009] Figure 1 A flowchart of a closed-loop detection noise modeling and suppression method is provided for an embodiment; Figure 2 A schematic diagram of the feedthrough cancellation circuit model; Figure 3 A schematic diagram of the automatic gain control module and the phase-locked loop dual closed-loop noise transmission circuit model; Figure 4 The following is an overall logic block diagram of a closed-loop detection noise modeling and suppression method provided for an embodiment. Detailed Implementation

[0010] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0011] like Figure 1 As shown, in one embodiment of the present invention, a closed-loop detection noise modeling and suppression method includes the following steps: S1. Construct the equivalent circuit model of the voltage sensor, the basic model of the automatic gain control module and the phase-locked loop dual closed-loop circuit; S2. Build a feedthrough cancellation circuit model and connect it to the equivalent circuit model of the voltage sensor to remove power supply noise; like Figure 2 As shown, the feedthrough cancellation circuit model includes: a driving electrode, a detection electrode, a resistor R1, a variable resistor R2, and a resistor R. m Operational amplifier U1, inductor L m Capacitor C1, Capacitor C f Capacitor C m The driving electrodes are connected in sequence to resistors R. m Inductor L m Capacitor C m and detection electrodes; capacitor C f One end is connected to one end of resistor R1 and one end of resistor R. m One end; capacitor C f The other end is connected to one end of capacitor C1 and capacitor C, respectively. m One end of the resistor is connected to pin 3 of the operational amplifier U1 and the other end of the variable resistor R2. Pin 1 of the operational amplifier U1 is grounded, and pin 4 of the operational amplifier U1 is connected to the sliding end of the variable resistor R2, the other fixed end of the variable resistor R2, and the other end of the capacitor C1.

[0012] S3. The electric field signal is detected by the sensing electrode of the equivalent circuit model of the voltage sensor and converted into a current signal; S4. The current signal is converted into a voltage signal through the induction electrode IV; S5. Filter the voltage signal as the input of the first bandpass filter to obtain the filtered voltage signal; S6. Use a phase-locked loop and an automatic gain control module to perform phase control and amplitude modulation on the filtered voltage signal, respectively. S7. Feed the signal after phase control and amplitude modulation back to the second bandpass filter for filtering to remove noise introduced by the phase-locked loop and automatic gain control module; The noise introduced by the phase-locked loop and automatic gain control module is determined by establishing a noise model and a noise transmission loop model; The phase-locked loop noise propagation path model includes the following sub-modules: Frequency and phase detectors / charge pumps (whose phase noise originates from internal circuit component noise), loop filters (used to filter out high-frequency components, whose phase noise mainly comes from resistor thermal noise and operational amplifier noise, such as voltage noise and current noise, which are converted into phase noise through the filter transfer function), voltage-controlled oscillators (the core frequency generation unit, whose phase noise is determined by its own noise characteristics), and frequency dividers (to achieve frequency division, whose phase noise is related to the frequency divider circuit structure); In the locked state, each submodule of the phase-locked loop operates within its own linear operating range. When the loop bandwidth is much smaller than the reference clock frequency, the phase-locked loop can be approximated as a linear continuous model for loop characteristic analysis. Based on linear system theory, the output phase noise from each noise source is derived.

[0013] The noise transfer function of the frequency and phase detector / charge pump is:

[0014] in, This is the noise transfer function of the frequency and phase detector / charge pump, reflecting its noise amplification and filtering characteristics after loop processing. It is the frequency division ratio. This represents the open-loop transfer function of a phase-locked loop; Pi Indicates the current of the charge pump; The noise transfer function of the loop filter is:

[0015] in, This is the noise transfer function of the loop filter, reflecting the noise after its own filtering and the effects of the loop. Indicates voltage-controlled gain. Represents the complex frequency variable in the Laplace transform; The noise transfer function of the voltage-controlled oscillator is:

[0016] in, This is the noise transfer function of the voltage-controlled oscillator, which directly affects the output. The noise transfer function of the frequency divider is:

[0017] in, The noise transfer function of the frequency divider; Since the noise sources are independent of each other, the total output phase noise is equal to the algebraic sum of the squares of the transfer function magnitudes of each noise source in the module. Therefore, the phase noise of the phase-locked loop output can be calculated as follows:

[0018] in, This refers to the phase noise output of the phase-locked loop. As a reference noise transfer function, For external reference phase noise, For the phase noise of the frequency-phase detector / charge pump, For the phase noise of the loop filter, This refers to the phase noise of the voltage-controlled oscillator. The phase noise of the frequency divider. This indicates calculating the square of the modulus.

[0019] After deriving the noise path transfer function, circuit models of each core module of AGC (Automatic Gain Control) and PLL (Phase-Locked Loop) are constructed using Cadence software. Module-level noise simulation is performed: noise simulations are conducted for PFD, CP, etc. (the simulation type is set to "Noise", and the frequency range covers the system's operating frequency band). The output noise power spectral density of each module is extracted and exported as a .csv file for subsequent import into the loop for path simulation.

[0020] The automatic gain control module and phase-locked loop dual closed-loop noise transmission loop model are as follows: Figure 3 As shown, where n p The noise introduced into the PLL loop, i.e., the phase noise of the phase-locked loop output. n m For mechanical thermal noise, n d For the path noise feedback at the AGC output, n e The noise at the circuit output is caused by each noise source being coupled to the system output through the transfer function of the corresponding loop module via the forward / feedback path, ultimately affecting the signal-to-noise ratio of the electric field detection signal.

[0021] The low-pass filter in the feedback loop prevents noise near the resonant frequency from passing through; therefore, the propagation path of mechanical and thermal noise to the resonator output can be considered open-loop. At frequencies deviating from the resonant frequency... The power spectrum at the output terminal is :

[0022] in, This refers to the noise spectrum of mechanical thermal noise itself. This is the noise transfer function of the equivalent sensor; Output circuit noise ne Including white noise and 1 / f Noise is controlled by the fact that the transfer functions of both the low-pass filter and the resonator are in the feedback loop, allowing only low-frequency noise to pass through while high-frequency noise is directly output.

[0023] in, V e Represents white noise. Indicates the frequency of transitions; The feedback loop consists only of the linearized sensor, low-pass filter, and PI controller. Let the linearized sensing function of the sensor be... ), n e The closed-loop transfer function can be written as:

[0024] in, This is the transfer function of the PI controller. Let be the transfer function of the LF low-pass filter; Therefore, the power spectral density of the closed-loop output of low-frequency noise is as follows:

[0025] The power spectrum can be obtained by calculating the correlation function of the noise and then performing a Fourier transform on the correlation function. Therefore, the power spectrum of the final output signal of the AGC loop is: :

[0026] in, The power spectrum of the modulated signal; The noise at the output of a phase-locked loop in linear systems can also be written as:

[0027] in, This represents the noise output of the phase-locked loop.

[0028] S8. Input the output of the second bandpass filter to the driving electrode of the equivalent circuit model of the voltage sensor and couple it to the output of the equivalent circuit model of the voltage sensor. Both the first bandpass filter and the second bandpass filter include: a signal frequency measurement module, a comparator, a frequency counting module, and a dynamically adjustable bandpass filter; The signal frequency measurement module is used to measure the frequency of the input signal; A comparator is used to convert the measured signal into a square wave signal of the same frequency; The frequency counting module is used to count the frequency of the square wave signal; The dynamically adjustable bandpass filter is used to compare the frequency measured by the frequency counting module with the center frequency of the dynamically adjustable bandpass filter. If the frequency fluctuation is within the preset frequency fluctuation threshold, the center frequency of the dynamically adjustable bandpass filter is adjusted; otherwise, the parameters of the dynamically adjustable bandpass filter are locked.

[0029] S9. Adaptive filtering of the output of the equivalent circuit model of the voltage sensor after IV conversion is performed using time-series-based adaptive Kalman filtering to obtain the noise-reduced voltage signal.

[0030] The specific method for adaptive filtering is as follows: The output of the equivalent circuit model of the voltage sensor is normalized (to eliminate the influence of dimensions) and its stationarity is checked (by processing non-stationary signals through differential operations) to obtain preprocessed data; Based on the preprocessed data, the order (p,q) of the ARMA model is determined according to the Akaike Information Criterion (AIC), and the autoregressive coefficients (AR) and moving average coefficients are estimated using the Yule-Walker equation to establish the ARMA model. The ARMA model is converted into a state-space form, and the state transition matrix and observation matrix are constructed. Adaptive filtering implementation: The Sage-Husa adaptive algorithm is used to estimate the process noise covariance Q and observation noise covariance R in real time based on the state transition matrix and observation matrix. Phase noise is dynamically filtered out through a "prediction-update" iteration (calculation of Kalman gain Kn, update of state estimate), outputting a denoised voltage signal. The updated state estimate is then used to obtain the denoised voltage signal. The overall logic block diagram of this invention is as follows: Figure 4 As shown, the signal after Kalman filtering is then subjected to "switched phase-sensitive detection" and "active filtering" to extract the DC component.

[0031] In summary, this invention optimizes hardware layout and closed-loop control from the source, designs dynamic adjustment capabilities based on power integrity, adapts to data processing under sudden load changes, forms a collaborative optimization architecture, and improves the system's overall ability to suppress phase noise. By constructing a complete noise suppression system and combining feedthrough capacitor active compensation technology with intelligent data processing algorithms, the sensor output signal is corrected and optimized in real time, achieving high-precision, low-noise voltage measurement.

Claims

1. A method for modeling and suppressing closed-loop detection noise, characterized in that, include: Construct the equivalent circuit model of the voltage sensor, the basic model of the automatic gain control module and the phase-locked loop dual closed-loop circuit; A feedthrough cancellation circuit model was built and connected to the equivalent circuit model of the voltage sensor to remove power supply noise; The electric field signal is detected by the sensing electrode of the equivalent circuit model of the voltage sensor and converted into a current signal. The current signal is converted into a voltage signal through the inductive electrode IV conversion module; The voltage signal is used as the input of the first bandpass filter for filtering to obtain the filtered voltage signal. Phase-locked loop (PLL) and automatic gain control (AGC) modules are used to perform phase control and amplitude modulation on the filtered voltage signal, respectively. The signal after phase control and amplitude modulation is fed back to the second bandpass filter for filtering to remove noise introduced by the phase-locked loop and automatic gain control module; The output of the second bandpass filter is input to the driving electrode of the equivalent circuit model of the voltage sensor and coupled to the output of the equivalent circuit model of the voltage sensor. The output of the coupled voltage sensor equivalent circuit model is adaptively filtered using a time-series-based adaptive Kalman filter to obtain a denoised voltage signal.

2. The method according to claim 1, characterized in that, The feedthrough cancellation circuit model includes: driving electrode, detection electrode, resistor R1, variable resistor R2, and resistor R. m Operational amplifier U1, inductor L m Capacitor C1, Capacitor C f Capacitor C m The driving electrodes are connected in sequence to resistors R. m Inductor L m Capacitor C m and detection electrodes; capacitor C f One end is connected to one end of resistor R1 and one end of resistor R. m One end; capacitor C f The other end is connected to one end of capacitor C1 and capacitor C, respectively. m One end of the resistor is connected to pin 3 of the operational amplifier U1 and the other end of the variable resistor R2. Pin 1 of the operational amplifier U1 is grounded, and pin 4 of the operational amplifier U1 is connected to the sliding end of the variable resistor R2, the other fixed end of the variable resistor R2, and the other end of the capacitor C1.

3. The method according to claim 1, characterized in that, The noise introduced by the phase-locked loop and automatic gain control module is determined by establishing a noise model and a noise transmission loop model; The phase-locked loop noise propagation path model includes the following sub-modules: Frequency and phase detectors / charge pumps, loop filters, voltage-controlled oscillators, and frequency dividers; The noise transfer function of the frequency and phase detector / charge pump is: in, Let be the noise transfer function of the frequency-phase detector / charge pump. It is the frequency division ratio. This represents the open-loop transfer function of a phase-locked loop; Pi Indicates the current of the charge pump; The noise transfer function of the loop filter is: in, Let be the noise transfer function of the loop filter. Indicates voltage-controlled gain. Represents the complex frequency variable in the Laplace transform; The noise transfer function of the voltage-controlled oscillator is: in, Let be the noise transfer function of the voltage-controlled oscillator; The noise transfer function of the frequency divider is: in, The noise transfer function of the frequency divider; The phase noise of the phase-locked loop output is: in, This refers to the phase noise output of the phase-locked loop. As a reference noise transfer function, For external reference phase noise, For the phase noise of the frequency-phase detector / charge pump, The phase noise of the loop filter, This refers to the phase noise of the voltage-controlled oscillator. The phase noise of the frequency divider. This indicates calculating the square of the modulus.

4. The method according to claim 3, characterized in that, The specific method for obtaining the noise of each sub-module in the noise transmission path loop model is as follows: Based on the noise transmission path loop model, the circuit models of the automatic gain control module and each sub-module in the phase-locked loop are constructed using simulation software, sub-module-level noise simulation is performed, and the output noise power spectral density of each sub-module is extracted.

5. The method according to claim 4, characterized in that, The simulation software is Cadence, the simulation type is "Noise", and the frequency range covers the system's operating frequency band.

6. The method according to claim 4, characterized in that, The expression for the output noise power spectral density is: in, This represents the output noise power spectral density. Represents the input noise spectrum; The noise transfer function of the equivalent sensor; The transformation gain of the electric field current This represents the gain of the current-voltage conversion.

7. The method according to claim 1, characterized in that, Both the first bandpass filter and the second bandpass filter include: a signal frequency measurement module, a comparator, a frequency counting module, and a dynamically adjustable bandpass filter; The signal frequency measurement module is used to measure the frequency of the input signal; A comparator is used to convert the measured signal into a square wave signal of the same frequency; The frequency counting module is used to count the frequency of the square wave signal; The dynamically adjustable bandpass filter is used to compare the frequency measured by the frequency counting module with the center frequency of the dynamically adjustable bandpass filter. If the frequency fluctuation is within the preset frequency fluctuation threshold, the center frequency of the dynamically adjustable bandpass filter is adjusted; otherwise, the parameters of the dynamically adjustable bandpass filter are locked.

8. The method according to claim 1, characterized in that, The specific method for adaptive filtering is as follows: The output of the equivalent circuit model of the voltage sensor is normalized and its stationarity is checked to obtain the preprocessed data. Based on the preprocessed data, the order of the ARMA model is determined according to the Akaike Information Criterion, and the autoregressive coefficients and moving average coefficients are estimated to establish the ARMA model. The ARMA model is converted into a state-space form, and the state transition matrix and observation matrix are constructed. The state estimate is estimated and updated in real time based on the state transition matrix and observation matrix; The denoised voltage signal is obtained using the updated state estimate.

Images