Intelligent electromagnetic interference suppression system for semiconductor equipment

Abstract

The invention relates to the technical field of semiconductor equipment, in particular to an intelligent electromagnetic interference suppression system for semiconductor equipment. The system comprises a sensor network which is installed at an electromagnetic interference key position of semiconductor equipment, monitors and collects electromagnetic data of the semiconductor equipment in real time, and sends the electromagnetic data to a central data processing module; the central data processing module is used for analyzing the received electromagnetic data, identifying and positioning an interference source based on an interference source positioning model of a machine learning algorithm, performing behavior prediction analysis on the interference source, generating a self-adaptive interference suppression strategy and sending an interference suppression instruction; and the feedback control module is used for executing a feedback control action based on the interference suppression instruction, realizing real-time suppression of multi-level electromagnetic interference of the semiconductor equipment and feeding back a suppression effect. According to the invention, real-time effective suppression of electromagnetic interference is realized, and the process stability and the control precision of semiconductor equipment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor equipment, and more specifically, to an intelligent electromagnetic interference suppression system for semiconductor equipment. Background Art

[0002] With the rapid development of semiconductor technology, the requirements of semiconductor equipment for electromagnetic environment are also increasing. Traditional electromagnetic interference suppression technologies, such as static shielding, filtering and grounding, can no longer meet the needs of modern semiconductor thin film equipment for stable operation in complex and changing electromagnetic environments.

[0003] The electromagnetic interference problems faced by semiconductor thin film equipment mainly include: high-frequency plasma interference, whose radiation energy affects signal stability and accuracy; electrode and power supply electromagnetic interference, which may interfere with sensitive electronic components and control systems, resulting in unstable operation; thermal noise and electromagnetic noise generated by high power density, which may be transmitted through wires or cables and affect other systems; and interference in the power supply and control system, which may affect power supply stability and control accuracy, resulting in process instability and affecting film quality and uniformity.

[0004] These electromagnetic interference problems will not only affect the stability and accuracy of the electronic system inside the equipment, but may also lead to instability in the process and a decrease in control accuracy, ultimately affecting product quality and production efficiency.

[0005] Therefore, there is an urgent need for an intelligent electromagnetic interference suppression system for semiconductor equipment to achieve accurate identification and effective suppression of electromagnetic interference in semiconductor equipment. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent electromagnetic interference suppression system for semiconductor equipment to solve the problem that the electromagnetic interference of semiconductor equipment is difficult to identify and suppress in the prior art.

[0007] In order to achieve the above object, the present invention provides a semiconductor device intelligent electromagnetic interference suppression system, including a sensor network, a central data processing module and a feedback control module:

[0008] The sensor network is connected to the central data processing module and installed at the key position of electromagnetic interference of the semiconductor device to monitor and collect electromagnetic data of the semiconductor device in real time and send it to the central data processing module;

[0009] The central data processing module is connected to the feedback control module, analyzes the electromagnetic data received from the sensor network, identifies and locates the interference source based on the interference source location model of the machine learning algorithm, performs behavior prediction analysis on the interference source, generates an adaptive interference suppression strategy, and sends an interference suppression instruction to the feedback control module;

[0010] The feedback control module executes feedback control actions based on the received interference suppression instructions to achieve real-time suppression of multi-level electromagnetic interference of semiconductor equipment and feedback suppression effects.

[0011] In some embodiments, the electromagnetic interference critical positions where the sensor network is installed include:

[0012] Power module, plasma chamber, control system and signal transmission channel.

[0013] In some embodiments, the sensor network collects electromagnetic waveform, spectrum and electric field strength data of semiconductor equipment in real time, and monitors the electromagnetic environment change data of semiconductor equipment at different process stages.

[0014] In some embodiments, the central data processing module includes a data preprocessing unit, a machine learning algorithm unit, and an adaptive optimization algorithm unit:

[0015] The data preprocessing unit is used to receive electromagnetic data from the sensor network and perform preprocessing and feature extraction;

[0016] The machine learning algorithm unit is used to dynamically model the interference source positioning model based on the machine learning algorithm and to identify and locate the interference source;

[0017] The adaptive optimization algorithm unit performs behavior prediction analysis on the interference source, generates an interference suppression strategy based on the intelligent control algorithm, and sends an interference suppression instruction to the feedback control module.

[0018] In some embodiments, the data preprocessing unit performs multi-level classification on the electromagnetic data received from the sensor network, and extracts high-frequency features, process timing features, and power matching features from the electromagnetic data.

[0019] In some embodiments, the data preprocessing unit performs feature extraction by the following steps:

[0020] Signal preprocessing step, removing noise and interference clutter from the received original electromagnetic signal;

[0021] A spectrum feature extraction step, extracting the unique frequency components of semiconductor devices based on a spectrum analysis method;

[0022] A time domain feature extraction step, extracting time-varying characteristic parameters of electromagnetic signals of semiconductor devices within a specific time period based on a time domain analysis method;

[0023] The spatial distribution feature extraction step extracts the electromagnetic field distribution features of different regions of the semiconductor device through spatial data fusion of a multi-sensor network;

[0024] The interference behavior pattern extraction step extracts the behavior pattern of a specific interference source in the semiconductor device.

[0025] In some embodiments, the machine learning algorithm unit analyzes a specific interference source and updates the interference source location model in real time as the high-frequency electromagnetic field in the semiconductor device changes with the process status.

[0026] In some embodiments, the specific interference sources include plasma electrode interference sources and power noise interference sources.

[0027] In some embodiments, the machine learning algorithm unit identifies and locates interference sources through the following steps:

[0028] Data labeling step: labeling the signal strength and phase information of the preprocessed data and synchronizing the multi-sensor data;

[0029] Model training and establishment steps: using supervised learning algorithm to train and establish interference source location model, and perform interference source identification;

[0030] In the interference source positioning step, the phase difference and signal strength difference data are input into the interference source positioning model for calculation to realize the positioning of the interference source.

[0031] In some embodiments, the supervised learning algorithm includes a support vector machine algorithm, a neural network algorithm, a random forest algorithm, and a gradient boosting tree algorithm.

[0032] In some embodiments, the adaptive optimization algorithm unit performs time series analysis prediction, process specific interference prediction, and equipment adaptive capability prediction based on an intelligent control algorithm.

[0033] In some embodiments, the adaptive optimization algorithm unit uses a regression model to perform time series analysis and prediction on interference sources.

[0034] In some embodiments, the regression model includes an autoregressive integrated moving average model and a long short-term memory network model.

[0035] In some embodiments, the adaptive optimization algorithm unit identifies the behavior pattern of the interference source by analyzing the interference signal characteristics in the historical data and the real-time data, and predicts the future performance of the interference source according to different operating conditions.

[0036] In some embodiments, the behavior pattern of the interference source includes periodic behavior, sudden behavior, and random behavior;

[0037] The adaptive optimization algorithm unit further recognizes the behavior pattern of the interference source, including:

[0038] The feature vectors of interference sources at different process stages are extracted, and the behavioral patterns of interference sources of semiconductor equipment are identified using a supervised learning model.

[0039] In some embodiments, the intelligent control algorithm of the adaptive optimization algorithm unit uses an adaptive interference suppression algorithm and a multi-level feedback control algorithm to generate an interference suppression strategy:

[0040] The adaptive interference suppression algorithm dynamically adjusts interference suppression parameters according to real-time data of different process stages, combined with the behavior pattern of the interference source and the historical data model;

[0041] The multi-level feedback control algorithm controls various subsystems of the semiconductor device in real time through a multi-level feedback loop.

[0042] In some embodiments, the intelligent control algorithm of the adaptive optimization algorithm unit further includes dynamic adjustment after interference source identification, intelligent filter self-adjustment mechanism, and predictive control and pre-adjustment.

[0043] In some embodiments, the feedback control module includes an intelligent control unit, an electromagnetic shielding control unit, an adaptive filter unit, and a multi-level suppression structure:

[0044] The intelligent control unit controls parameter variables of the semiconductor device in real time based on control instructions received from the central data processing module;

[0045] The electromagnetic shielding control unit dynamically adjusts the electromagnetic shielding material based on the control instructions received from the central data processing module;

[0046] The adaptive filter unit adjusts filter parameters of the semiconductor device based on control instructions received from the central data processing module;

[0047] The multi-level suppression structure combines multiple feedback loops of the semiconductor device to form a closed-loop control structure with step-by-step feedback, thereby achieving interference suppression of the semiconductor device under different process conditions.

[0048] In some embodiments, the multi-level suppression structure performs multi-level collaborative interference suppression on the interference source layer, the transmission path layer, and the receiving end layer.

[0049] The present invention provides an intelligent electromagnetic interference suppression system for semiconductor equipment, which uses a machine learning model to identify interference sources and predict behaviors, and an intelligent control algorithm to dynamically adjust electromagnetic interference suppression measures. The multi-level suppression structure effectively controls electromagnetic interference from different levels, thereby achieving real-time and effective suppression of electromagnetic interference, and significantly improving the process stability and control accuracy of semiconductor equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The above and other features, properties and advantages of the present invention will become more apparent through the following description in conjunction with the accompanying drawings and embodiments, in which the same reference numerals always represent the same features, wherein:

[0051] Figure 1 A principle block diagram of an intelligent electromagnetic interference suppression system for semiconductor equipment according to an embodiment of the present invention is disclosed;

[0052] Figure 2 An architecture diagram of an intelligent electromagnetic interference suppression system for semiconductor equipment according to an embodiment of the present invention is disclosed;

[0053] Figure 3 A principle block diagram of a central data processing module according to an embodiment of the present invention is disclosed;

[0054] Figure 4 A schematic diagram of time domain and frequency domain analysis results of electromagnetic data according to an embodiment of the present invention is disclosed;

[0055] Figure 5 A schematic diagram of prediction results of an LSTM model according to an embodiment of the present invention is disclosed;

[0056] Figure 6 A schematic diagram of interference suppression effect of an intelligent control algorithm according to an embodiment of the present invention is disclosed;

[0057] Figure 7 A principle block diagram of a feedback control module according to an embodiment of the present invention is disclosed.

[0058] The meanings of the reference numerals in the figures are as follows:

[0059] 10Sensor network;

[0060] 20 central data processing module;

[0061] 21 data preprocessing unit;

[0062] 22 machine learning algorithm units;

[0063] 23 adaptive optimization algorithm unit;

[0064] 30 feedback control module;

[0065] 31 intelligent control unit;

[0066] 32 electromagnetic shielding control unit;

[0067] 33 adaptive filter unit;

[0068] 34Multi-level inhibition structure. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the invention and are not used to limit the invention.

[0070] Figure 1 The principle block diagram of the intelligent electromagnetic interference suppression system for semiconductor devices according to an embodiment of the present invention is disclosed. Figure 1 As shown, a semiconductor device intelligent electromagnetic interference suppression system proposed by the present invention includes a sensor network 10, a central data processing module 20 and a feedback control module 30:

[0071] The sensor network 10 is connected to the central data processing module 20 and installed at the key position of electromagnetic interference of the semiconductor device to monitor and collect electromagnetic data of the semiconductor device in real time and send it to the central data processing module 20;

[0072] The central data processing module 20 is connected to the feedback control module 30, analyzes the electromagnetic data sent by the received sensor network 10, identifies and locates the interference source based on the interference source location model of the machine learning algorithm, performs behavior prediction analysis on the interference source, generates an adaptive interference suppression strategy, and sends an interference suppression instruction to the feedback control module 30;

[0073] The feedback control module 30 executes feedback control actions based on the received interference suppression instructions to achieve real-time suppression of multi-level electromagnetic interference on semiconductor devices and feedback the suppression effect.

[0074] The intelligent electromagnetic interference (EMI) suppression system proposed in the present invention can be applied to semiconductor PECVD (plasma enhanced chemical vapor deposition) thin film equipment, including a sensor network, a central data processing module, and a feedback control module. The modules are connected through a high-speed data bus to form a closed-loop feedback control system to achieve real-time electromagnetic interference monitoring, intelligent analysis and adaptive interference suppression.

[0075] Figure 2 The architecture diagram of the intelligent electromagnetic interference suppression system for semiconductor devices according to an embodiment of the present invention is disclosed. Figure 1 and Figure 2 , the various component modules of the semiconductor device intelligent electromagnetic interference suppression system proposed in the present invention will be described in detail.

[0076] The sensor network 10 collects electromagnetic waveform, spectrum and electric field strength data of semiconductor equipment in real time, and monitors the electromagnetic environment change data of semiconductor equipment at different process stages.

[0077] The sensor network 10 is responsible for collecting data on electromagnetic waveforms, spectrum, and electric field strength in semiconductor devices in real time, and monitoring changes in the electromagnetic environment of the device at different process stages, including plasma ignition, stabilization, and shutdown. The sensor network 10 provides accurate data support for the central data processing module 20 to analyze and locate interference sources.

[0078] Since the space inside the PECVD thin film equipment is relatively compact, the sensor layout needs to be precisely designed to ensure that all possible locations of interference sources are covered and to avoid interference between sensors. On this basis, the sensor network 10 also needs to be differently arranged inside and outside the plasma chamber in order to obtain more accurate interference location data.

[0079] The layout principle of the sensor network 10 is to ensure full coverage first, that is, the sensors are evenly distributed inside key components and around the equipment to cover the electromagnetic environment. In key areas, the focus of sensor layout is RF power supply, RPS power supply, equipment power line, cable connection, equipment interface and near possible interference sources. In order to ensure the accuracy of measurement data, the layout of the sensor network 10 also considers interference isolation to avoid mutual interference between sensors.

[0080] In this embodiment, the sensor network 10 is installed at the key positions of electromagnetic interference, including the power module, the plasma chamber, the control system and the signal transmission channel:

[0081] Location near the power module: used to monitor the electromagnetic radiation and interference waveform generated by the high-frequency power supply to ensure the stability of the power supply under different load conditions.

[0082] Plasma chamber: Sensors are installed around the plasma electrodes and reaction chamber to monitor the high-frequency electromagnetic field generated by the plasma in real time and capture instantaneous electromagnetic wave changes.

[0083] Around the control system: used to detect the electromagnetic environment near the control system to prevent interference signals from affecting the precise control of the system.

[0084] Signal transmission channel: Sensors are placed near the signal lines and power lines in the equipment to monitor electromagnetic interference on the transmission lines and prevent interference from propagating through the wires.

[0085] In order to achieve high-sensitivity electromagnetic environment monitoring, the sensor network 10 selects high-frequency electromagnetic sensors as measurement tools, such as ETS-Lindgren 7405, Tektronix CT-6, Aaronia PBS1 and other models. These sensors can detect electromagnetic signals from several kHz to several GHz, meeting the requirements of the operating frequency range of semiconductor manufacturing equipment.

[0086] The central data processing module 20 is responsible for the collection, processing and analysis of electromagnetic data, using advanced machine learning algorithms to model and identify interference sources, and develop optimized suppression strategies. By analyzing the data of semiconductor equipment at different process stages, the most destructive electromagnetic interference sources can be identified. In addition, the interference source location model of the equipment can be dynamically updated, and an adaptive interference suppression strategy can be generated based on this, and an interference source behavior prediction and analysis function can be provided, which can sense possible interference in the equipment in advance, thereby controlling the feedback control module 30 to take preventive interference suppression measures.

[0087] Figure 3 The principle block diagram of the central data processing module according to an embodiment of the present invention is disclosed. Figure 3 As shown, further, the central data processing module 20 includes a data preprocessing unit 21, a machine learning algorithm unit 22 and an adaptive optimization algorithm unit 23:

[0088] The data preprocessing unit 21 is used to receive electromagnetic data from the sensor network 10 and perform preprocessing (such as denoising and signal enhancement) and feature extraction;

[0089] The machine learning algorithm unit 22 is used to dynamically model the interference source positioning model based on the machine learning algorithm and identify and locate the interference source;

[0090] The adaptive optimization algorithm unit 23 performs behavioral prediction analysis on the interference source, generates an interference suppression strategy based on the intelligent control algorithm, and sends an interference suppression instruction to the feedback control module 30, thereby adjusting the power supply parameters, shielding layer configuration and filter parameters of the device in real time.

[0091] Furthermore, the data preprocessing unit 21 performs multi-level classification on the electromagnetic data received from the sensor network 10, and extracts high-frequency features, process timing features, and power matching features from the electromagnetic data.

[0092] The function of the feature extraction performed by the data preprocessing unit 21 is to analyze the extremely characteristic electromagnetic interference signal in the semiconductor device and quantify the electromagnetic environment of the device at different process stages.

[0093] Due to the presence of extremely strong high-frequency electromagnetic fields in semiconductor devices, especially electromagnetic waveforms from plasma power supplies, the interference characteristics change significantly with the generation and disappearance of plasma. Multi-level classification of electromagnetic data of sensor network 10 refers to multi-level sensor classification to collect interference data from different areas, such as instantaneous high-frequency interference in the plasma chamber, low-frequency noise of the power input module, etc.

[0094] For PECVD thin film equipment, the high-frequency features in its electromagnetic data refer to the sharp spectrum features generated by the plasma electrode excitation unique to PECVD thin film equipment, which is closely related to the excitation frequency and energy of the plasma. The process timing features in the electromagnetic data refer to the significant differences in the electromagnetic interference of PECVD thin film equipment during the startup, stabilization, and shutdown stages of the process. Extracting these time series features helps to monitor and predict interference in real time. The power matching features in the electromagnetic data refer to the extraction of the complex electromagnetic waveform features generated by the power matching circuit of the PECVD thin film equipment in response to the dynamic changes of the electromagnetic field under high power density.

[0095] More specifically, for semiconductor devices, especially PECVD thin film devices, the data preprocessing unit 21 performs feature extraction through the following steps:

[0096] The signal preprocessing step is to remove noise and interference from the received original electromagnetic signal.

[0097] Since the working environment of the PECVD thin film equipment contains a large number of unique instantaneous high-frequency interferences (such as spike pulses during plasma ignition), the above steps use bandpass filters and wavelet transforms to pre-process the signals.

[0098] The spectrum feature extraction step extracts the unique frequency components of semiconductor devices based on the spectrum analysis method, especially the harmonic distribution in the high-frequency plasma excitation process.

[0099] Spectrum analysis methods include fast Fourier transform (FFT) and others. The high-frequency harmonic features extracted in the above steps can accurately indicate the propagation of electromagnetic waves generated when plasma is excited. In view of the spectrum characteristics of PECVD thin film equipment, the system pays special attention to high-frequency interference with excitation frequencies in the GHz range. This interference source generally does not appear in general equipment and is a significant feature of the PECVD process.

[0100] The time domain feature extraction step extracts the time-varying characteristic parameters of the electromagnetic signal of the semiconductor device within a specific time period based on the time domain analysis method, especially the intensity, duration and repetition frequency of the interference signal.

[0101] Time domain analysis methods include autocorrelation function and short-time Fourier transform. In time domain analysis, the system specifically extracts the spike interference signal synchronized with plasma excitation and analyzes its impact on temperature fluctuation and power matching state during film deposition. These features can only be obtained through the unique timing mode analysis of PECVD thin film equipment and cannot be directly applied in general equipment.

[0102] The spatial distribution feature extraction step extracts the electromagnetic field distribution features of different areas of the semiconductor device through spatial data fusion of a multi-sensor network.

[0103] The key information extracted in the above steps includes the distribution of electromagnetic fields inside and outside the plasma chamber, with special attention paid to the impact of high-frequency electric fields on surrounding electronic systems and process control modules. These spatial distribution characteristics directly affect the uniformity and deposition rate of PECVD films, especially in the complex electromagnetic environment of process chamber design. This analysis method is unique to PECVD thin film equipment.

[0104] The interference behavior pattern extraction step extracts the behavior pattern of specific interference sources in semiconductor devices, including the typical evolution trajectory and duration of the interference signal at different stages of the process.

[0105] Based on these behavior patterns, the system can predict the interference that may occur in the next stage in advance, which is convenient for real-time adjustment in the feedback control module. In PECVD thin film equipment, the interference behavior generated by plasma electrodes and power modules has obvious periodic and stage characteristics. The extracted behavior patterns pay special attention to the impact of these interferences on the thin film deposition process, as well as the real-time changes in temperature, pressure, and plasma state inside the equipment. These behavior patterns reflect the unique electromagnetic interference characteristics of PECVD thin film equipment under different process conditions.

[0106] Figure 4 The present invention discloses a schematic diagram of the time domain and frequency domain analysis results of electromagnetic data according to an embodiment of the present invention. Figure 4 In the diagram on the left, the time domain analysis clearly depicts the waveform of the signal over time and intuitively shows the dynamic changes of the signal amplitude over time. Figure 4 On the right, the frequency domain analysis part relies on the fast Fourier transform (FFT) technology to achieve accurate conversion of the signal from the time domain to the frequency domain. The diagram on the right clearly shows the frequency composition of the signal, and clearly reveals that the signal is mainly composed of two frequency components, 50Hz and 120Hz. The above analysis method has shown extremely high application value in electromagnetic compatibility (EMC) research and can effectively identify and analyze the characteristics of electromagnetic interference.

[0107] The machine learning algorithm unit 22 analyzes specific interference sources and updates the interference source location model in real time according to the changes of high-frequency electromagnetic fields in semiconductor equipment as the process status changes.

[0108] The specific interference sources include plasma electrode interference sources and power noise interference sources.

[0109] Due to the extremely complex coupling relationship between the plasma electrode and the power system in the PECVD thin film equipment, the electrode will produce a tip effect when working, resulting in an excessively strong local electromagnetic field. By identifying the precise location and intensity of these plasma electrode interference sources, their impact on the uniformity of thin film deposition is analyzed.

[0110] When PECVD thin film equipment is in high power mode, the power matching system will introduce low-frequency oscillation interference. Identifying these plasma electrode interference sources helps maintain power supply stability during the process.

[0111] The PECVD process has multiple stages, and the performance of electromagnetic interference sources in different process stages varies significantly. Therefore, the interference source location algorithm must be able to adapt to the complex electromagnetic environment of the equipment in different states. This requires the model to have dynamic adjustment capabilities and be able to learn and update online according to the real-time process status. The interference source location model of the present invention adopts a dynamic modeling method, which specifically updates the model in real time for the high-frequency electromagnetic field in the PECVD thin film equipment as the process status changes. For example, according to different film materials, process gas ratios and power load changes, the interference source location model is updated to perform more accurate interference source location.

[0112] More specifically, for semiconductor equipment, especially PECVD thin film equipment, the machine learning algorithm unit 22 identifies and locates interference sources through the following steps:

[0113] Data labeling step: labeling the signal strength and phase information of the preprocessed data and synchronizing the multi-sensor data;

[0114] Model training and establishment steps: using supervised learning algorithm to train and establish interference source location model, and perform interference source identification;

[0115] In the interference source positioning step, the phase difference and signal strength difference data are input into the interference source positioning model for calculation to realize the positioning of the interference source.

[0116] More specifically, the electromagnetic interference characteristics generated by PECVD thin film equipment at different process stages vary significantly, so the collected data should cover interference source signals under multiple process conditions. For example, high-frequency spike signals in the plasma startup phase, continuous electromagnetic wave interference during thin film deposition, and power attenuation signals during shutdown. When annotating these data, they need to be classified according to different process stages so that the subsequent model can learn the interference characteristics under different states.

[0117] Among them, since the sensor network of the PECVD thin film equipment is installed in the plasma chamber, near the electrode and the equipment control module, the signal strength received by the sensor will vary with the distance and power of the interference source. Therefore, it is necessary to specially mark the intensity level of each signal and mark the signal strength so that the subsequent algorithm can determine the approximate location of the interference source based on the intensity difference.

[0118] In addition to signal strength, phase information is also crucial in locating interference sources. In PECVD thin film equipment, the interference waves of high-frequency plasma usually have the characteristics of phase changes. These phase differences reflect the propagation path and direction of the interference signal. For this reason, the phase of the signal received by each sensor must be accurately recorded during the data annotation process, which is very critical for subsequent positioning calculations based on phase differences.

[0119] In the multi-sensor network of PECVD thin film equipment, the signal propagation speed is extremely fast under the interference of high-frequency plasma, and even a small time difference will affect the positioning accuracy. Therefore, it is necessary to strictly control the sampling time when annotating data and ensure the synchronization of the time base of the data of multiple sensors.

[0120] Furthermore, the supervised learning algorithms in the model training establishment step include support vector machine algorithm, neural network algorithm, random forest algorithm and gradient boosting tree algorithm.

[0121] By comparing various supervised learning algorithms, it is found that ensemble learning methods (such as random forests and gradient boosting trees) are superior in locating electromagnetic interference sources in PECVD thin film equipment. This type of model can use a combination of multiple decision trees to process complex signal data, including intensity and phase information, and accurately locate interference sources through training with a large amount of labeled data.

[0122] Furthermore, the phase difference in the interference source location step is one of the key factors in locating the interference source. Through the phase difference between multiple sensors, the system can determine the propagation path of the interference signal and calculate the precise location of the interference source. This calculation based on phase difference is particularly prominent in high-frequency plasma interference environments, because high-frequency signals propagate quickly and phase difference changes are sensitive, which can provide higher positioning accuracy.

[0123] Furthermore, the signal strength in the interference source location step is used to further correct the deviation caused by the phase difference. By comparing the strength signals received by different sensors, the model can roughly estimate the distance of the interference source and combine the phase difference information to accurately locate the interference source.

[0124] Since the high-frequency plasma interference commonly found in PECVD thin-film equipment has an extremely high temporal and spatial rate of change, the tiny phase difference of the high-frequency signal must be accurately measured and calculated during the positioning process, and ordinary low-frequency signal processing methods are obviously insufficient here. In the experimental verification of PECVD thin-film equipment, the supervised learning model based on the combination of signal strength and phase difference controlled the error in the interference source positioning within a few centimeters, meeting the equipment's requirements for accurate positioning of the interference source.

[0125] More specifically, for semiconductor equipment, especially PECVD thin film equipment, the adaptive optimization algorithm unit 23 performs time series analysis prediction, process specific interference prediction and equipment adaptive capability prediction based on the intelligent control algorithm.

[0126] Time series analysis is used to identify and predict the time-varying trends of interference signals. In PECVD thin film equipment, due to factors such as plasma excitation and high-frequency power supply fluctuations, interference signals usually show strong time correlation, so using time series models for prediction can achieve better results.

[0127] Furthermore, the adaptive optimization algorithm unit 23 uses a regression model to perform time series analysis and prediction on interference sources.

[0128] The regression model includes an autoregressive integrated moving average (ARIMA) model and a long short-term memory network (LSTM) model.

[0129] The autoregressive integrated moving average (ARIMA) model predicts the future value of time series data by combining the autoregressive (AR) and moving average (MA) models. It is suitable for predicting relatively stable and linear interference signals, such as power supply fluctuations and plasma noise interference of PECVD thin film equipment under stable working conditions. This model can capture the law of periodic fluctuations of equipment in standard process flow and is used to predict interference changes in the short term.

[0130] Long short-term memory (LSTM) is a deep learning model based on recurrent neural network (RNN) that can process time series data with long-term dependencies. Compared with ARIMA, LSTM model is better at capturing complex, nonlinear time dependencies, making it particularly suitable for predicting interference signals caused by high-frequency plasma and complex electromagnetic environment in PECVD thin film equipment. LSTM can remember long historical data and capture the electromagnetic interference patterns of equipment at different process stages, so it is suitable for predicting non-periodic fluctuations of interference signals, especially sudden changes caused by process conversion (such as plasma excitation or shutdown process).

[0131] Figure 5 A schematic diagram of the prediction results of the LSTM model according to an embodiment of the present invention is disclosed, such as Figure 5 As shown in the figure, the LSTM model can follow the trend of the actual signal well. Although there are slight deviations at some time points, the overall trend is consistent.

[0132] Through experimental comparison, the LSTM model performs better in processing high-frequency plasma interference and complex dynamic signals of PECVD thin film equipment, and can better predict sudden fluctuations of signals, while the ARIMA model is more reliable when the equipment is in a relatively stable stage. Therefore, combining the advantages of both for hybrid modeling can improve prediction accuracy. Therefore, the ARIMA model is used in the stable process stage, and the LSTM model is used in the complex conversion stage.

[0133] The adaptive optimization algorithm unit 23 identifies the behavior pattern of the interference source by analyzing the interference signal characteristics in the historical data and the real-time data, and predicts the future performance of the interference source according to different operating conditions.

[0134] In PECVD thin film equipment, the behavior pattern of interference sources has a strong dependence on the equipment process, so pattern recognition and prediction are required for specific process conditions.

[0135] The behavior patterns of the interference sources include periodic behavior, sudden behavior and random behavior:

[0136] The periodic behavior, such as the interference phenomenon in a stable plasma state, has a relatively constant interference intensity and frequency, showing a periodic fluctuation characteristic.

[0137] During the sudden behaviors, such as plasma ignition or power transient, the interference signal exhibits short-term and violent fluctuations.

[0138] The random behavior is irregular noise caused by the device control system or external electromagnetic interference, which appears as random fluctuations.

[0139] The adaptive optimization algorithm unit further recognizes the behavior pattern of the interference source, including:

[0140] The feature vectors of interference sources at different process stages are extracted, and the behavioral patterns of interference sources of semiconductor equipment are identified using a supervised learning model.

[0141] Through features such as signal strength, frequency, and phase change, feature vectors of interference sources at different process stages are extracted. In supervised learning models, support vector machines (SVM) and random forests perform better in identifying interference behavior patterns of PECVD thin film equipment. SVM is good at processing high-dimensional feature data and is suitable for identifying complex interference behavior patterns; while random forests can effectively distinguish different patterns and predict their development trends by integrating multiple decision trees.

[0142] After identifying the behavior pattern, the system predicts the future performance of the interference source based on the current operating conditions (such as power, temperature, plasma state, etc.). For example, the system can predict whether the interference signal will fluctuate drastically when the equipment power increases, and take corresponding interference suppression measures in advance to prevent process interruption or equipment abnormality.

[0143] By mining historical data and continuously updating real-time data, the system can not only identify the current behavior patterns of interference sources, but also predict future interference situations based on the evolution trend of the patterns. This process is particularly effective in PECVD thin film equipment, because the process status of the equipment changes frequently, and the interference effects of sudden behaviors on the process can be identified and predicted in advance, thereby ensuring the stable operation of the equipment.

[0144] By combining ARIMA and LSTM models for time series analysis and using supervised learning models (SVM, random forest) for behavioral pattern recognition, the system can effectively predict the future performance of interference sources and take corresponding measures based on the prediction results. Therefore, it fully considers the process changes and complex electromagnetic environment in PECVD equipment, ensures the accuracy and pertinence of the prediction process, and avoids the limitations of general methods.

[0145] More specifically, for semiconductor equipment, especially PECVD thin film equipment, the intelligent control algorithm of the adaptive optimization algorithm unit 23 uses an adaptive interference suppression algorithm and a multi-level feedback control algorithm to generate an interference suppression strategy:

[0146] The adaptive interference suppression algorithm dynamically adjusts interference suppression parameters according to real-time data of different process stages, combined with the behavior pattern of the interference source and the historical data model;

[0147] The multi-level feedback control algorithm controls various subsystems of the semiconductor device in real time through a multi-level feedback loop.

[0148] PECVD equipment has highly complex interference sources during the process, such as nonlinear interaction between high-frequency power supply and plasma. The Adaptive Interference Mitigation Algorithm (AIMA) can dynamically adjust the interference suppression parameters based on real-time data from different process stages (such as plasma startup, film deposition, electrode voltage adjustment, etc.). This adaptive adjustment is based on real-time feedback from the system and combined with historical data models to ensure that the suppression strategy can respond quickly when the electromagnetic environment changes suddenly. For example, when the plasma electrode generates a strong electric field, the AIMA algorithm will automatically increase the electromagnetic shielding strength or adjust the frequency response range of the interference suppression filter.

[0149] The multi-level feedback control algorithm is divided into three main levels: micro-level feedback (signal feedback within the subsystem), meso-level feedback (coordinated control between system submodules), and macro-level feedback (overall control and state adjustment of the entire device).

[0150] More specifically, in PECVD equipment, the process is highly coupled with electromagnetic interference. The multi-level feedback control algorithm implements different suppression measures at each level according to the different behavior patterns of the electromagnetic interference source. For example, at the micro level, MLFC adjusts the shielding strength around a specific electrode according to the real-time electromagnetic state of the electrode; at the macro level, MLFC adjusts the overall power supply frequency or plasma power according to the process progress of the entire equipment to reduce the interference effect on sensitive subsystems.

[0151] The intelligent control algorithm of the adaptive optimization algorithm unit 23 also includes dynamic adjustment after interference source identification, intelligent filter self-adjustment mechanism, and predictive control and pre-adjustment.

[0152] The influence of plasma state on the characteristics of interference sources in PECVD equipment is relatively complex, especially when the equipment is in different process stages, the spectrum and intensity of interference change significantly. Intelligent control algorithms will adjust the suppression measures in real time based on these specific changes.

[0153] Specifically, the dynamic adjustment strategies after interference source identification include:

[0154] When a specific interference source is identified (such as plasma high-frequency interference or power supply transient noise), the intelligent control algorithm will dynamically adjust the parameters of the suppression device according to the characteristics of the interference source. For example, when a sudden interference from high-frequency plasma is identified, the intelligent control algorithm will adjust the frequency range of the bandpass filter to cover the main frequency band of the high-frequency noise and prevent the interference signal from entering the critical control circuit.

[0155] Electromagnetic interference in PECVD thin film equipment often has a complex spectrum distribution, especially when working with high-frequency plasma. Therefore, the intelligent control algorithm includes a self-adjusting filter mechanism that can dynamically adjust the filter's frequency response range based on real-time feedback data.

[0156] Specifically, the self-adjusting filter mechanism can adjust the bandwidth and center frequency of the filter according to real-time parameters such as plasma density and power input to cope with frequently changing high-frequency noise. For example, when high-frequency electromagnetic wave interference caused by power changes between electrodes occurs, the self-adjusting filter will expand the bandwidth range to a specific frequency band to minimize interference with other sensitive electronic systems.

[0157] The intelligent control algorithm also includes predictive control and pre-adjustment mechanisms. By combining the output of the behavior prediction module, the system parameters are pre-adjusted in advance to avoid the adverse effects of impending interference sources on the equipment.

[0158] Changes in the process state of PECVD equipment (such as instantaneous enhancement of plasma or increase in power) often cause sudden fluctuations in interference signals. Intelligent control algorithms use predictive models to learn about these interference changes in advance and make pre-adjustments. For example, when it is expected that plasma will be ignited, the algorithm will pre-adjust the frequency response of the power supply and the shielding strength of the interference shield to prevent high-frequency interference from affecting other subsystems of the equipment.

[0159] Figure 6 A schematic diagram of the interference suppression effect of an intelligent control algorithm according to an embodiment of the present invention is disclosed. The intelligent control algorithm relies on the results of data processing and analysis to dynamically adjust various measures including shielding, filtering, and signal isolation to achieve interference suppression measures. Figure 6 The comparison before and after the implementation of these measures is intuitively shown. By carefully analyzing the comparative data of the three experiments, it can be clearly concluded that:

[0160] Shielding measures: Figure 6 The top graph shows the significant change in electromagnetic radiation intensity before and after the application of shielding materials around the interference source. Specifically, after the shielding is implemented, the radiation intensity is significantly reduced by more than 85%.

[0161] Filtering measures: Figure 6 The middle diagram shows in detail the characteristics of the conducted interference signal changing with frequency after the introduction of the high-efficiency filter. The comparison results show that after the introduction of the filter, the amplitude of the interference signal in the key frequency band has dropped significantly by 70%.

[0162] Signal isolation: Figure 6 The bottom diagram shows the change in interference signal strength at the receiving end before and after the use of optical isolation technology. Experimental data shows that through the application of signal isolation technology, the interference strength at the receiving end has been effectively reduced by 68%.

[0163] These experimental data fully demonstrate the remarkable effectiveness of the intelligent control algorithm of the present invention in different interference suppression measures.

[0164] The intelligent control algorithm proposed in the present invention can effectively cope with complex and changeable electromagnetic environments and ensure the normal operation of the equipment by dynamically and adaptively adjusting interference suppression measures. It fully considers the process characteristics of PECVD equipment and the particularity of electromagnetic interference. Different from traditional static interference suppression methods, it can intelligently adjust the interference suppression strategy according to the real-time status of the equipment and the progress of the process to achieve precise control of interference.

[0165] The feedback control module 30 dynamically adjusts the operating parameters of the equipment according to the control instructions issued by the central data processing module, such as adjusting the frequency and power of the plasma, or changing the working mode of the power supply. The layout and form of the electromagnetic shielding material are controlled in real time to optimize the shielding effect and avoid the interference signal from affecting the core components of the equipment. The working frequency band and attenuation coefficient of the filter are dynamically adjusted to ensure that the filtering effect can be adaptively adjusted according to process changes.

[0166] Figure 7 A principle block diagram of a feedback control module according to an embodiment of the present invention is disclosed. Figure 7 As shown, more specifically, for semiconductor equipment, especially PECVD thin film equipment, the feedback control module 30 includes an intelligent control unit 31, an electromagnetic shielding control unit 32, an adaptive filter unit 33 and a multi-level suppression structure 34:

[0167] The intelligent control unit 31 controls parameter variables of the semiconductor device in real time based on the control instructions received from the central data processing module 20, wherein the parameter variables include power supply parameters, plasma operation parameters and other process variables;

[0168] The electromagnetic shielding control unit 32 dynamically adjusts the electromagnetic shielding material based on the control instructions received from the central data processing module 20 to achieve targeted interference isolation and shielding optimization;

[0169] The adaptive filter unit 33 adjusts the filter parameters of the semiconductor device, especially the filter parameters of the power supply system, based on the control instructions received from the central data processing module 20, to reduce electromagnetic interference in different frequency bands;

[0170] The multi-level suppression structure 34 combines multiple feedback loops of the semiconductor device to form a closed-loop control structure with step-by-step feedback, thereby achieving interference suppression of the semiconductor device under different process conditions.

[0171] The multi-level suppression structure performs multi-level collaborative interference suppression on the interference source layer, the transmission path layer and the receiving end layer.

[0172] The optimization design of the interference source layer is realized based on the previous interference identification and analysis algorithms (such as feature extraction and pattern recognition). This layer directly deploys physical suppression measures for interference sources, such as optimizing the shape and layout of plasma electrodes, improving the grounding system or shielding structure.

[0173] Compared with other general equipment, the optimized design of the interference source layer combines the unique process requirements of semiconductor PECVD thin film equipment. For example, the change of plasma power frequency and the electrode position have a particularly important impact on the quality of the film. Therefore, the optimized design of this layer is a special adjustment made for this unique high-frequency interference environment to ensure the maximum interference suppression effect.

[0174] Implement shielding and filtering strategies at the transmission path layer: Shielding strategies include shielding measures for cables and signal transmission lines, such as shielded cables and shielded boxes. Filtering strategies involve embedding filters in cables and signal paths to filter out high-frequency interference signals.

[0175] The implementation of anti-interference design at the receiving end layer includes two main aspects: the first is shielding design, that is, equipping the receiving end equipment with an effective shielding cover to reduce the impact of external interference signals. The second is anti-interference circuit design, by introducing anti-interference measures such as differential mode filters and common mode filters in the receiving circuit, the interference components in the received signal can be further reduced.

[0176] The semiconductor equipment intelligent electromagnetic interference suppression system proposed by the present invention has the following beneficial effects:

[0177] 1) Improve EMC performance: With real-time detection and dynamic control capabilities, it can effectively suppress electromagnetic interference, thereby significantly enhancing the electromagnetic compatibility performance of semiconductor equipment;

[0178] 2) Strong adaptability: Using advanced machine learning algorithms, it can automatically learn and optimize interference suppression strategies to adapt to different electromagnetic environments;

[0179] 3) Multi-level collaborative suppression: Collaborative suppression is performed at multiple levels, including interference source, transmission path, and receiving end, with significant suppression effect;

[0180] 4) Low power consumption and high performance: By optimizing the system design, the system can reduce power consumption as much as possible while ensuring high performance;

[0181] 5) Good system integration and compatibility: The system can be seamlessly integrated into existing equipment and maintain good compatibility, providing users with the convenience of seamless upgrades and expansions.

[0182] Those skilled in the art will appreciate that information, signals, and data may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips cited throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

[0183] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, units, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, boxes, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0184] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0185] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0186] The above embodiments are provided for persons familiar with the art to implement or use the present invention. Personnel familiar with the art can make various modifications or changes to the above embodiments without departing from the inventive concept of the present invention. Therefore, the protection scope of the present invention is not limited to the above embodiments, but should be the maximum scope of the innovative features mentioned in the claims.

Claims

1. A semiconductor equipment intelligent electromagnetic interference suppression system, characterized in that: Including sensor network, central data processing module and feedback control module: The sensor network is connected to the central data processing module and installed at the key position of electromagnetic interference of the semiconductor device to monitor and collect electromagnetic data of the semiconductor device in real time and send it to the central data processing module; The central data processing module is connected to the feedback control module, analyzes the electromagnetic data received from the sensor network, identifies and locates the interference source based on the interference source location model of the machine learning algorithm, performs behavior prediction analysis on the interference source, generates an adaptive interference suppression strategy, and sends an interference suppression instruction to the feedback control module; The feedback control module executes feedback control actions based on the received interference suppression instructions to achieve real-time suppression of multi-level electromagnetic interference of semiconductor equipment and feedback suppression effects.

2. The semiconductor device intelligent electromagnetic interference suppression system according to claim 1, characterized in that: The sensor network is installed at key locations for electromagnetic interference, including: Power module, plasma chamber, control system and signal transmission channel.

3. The semiconductor device intelligent electromagnetic interference suppression system according to claim 1, characterized in that: The sensor network collects electromagnetic waveform, spectrum and electric field strength data of semiconductor equipment in real time, and monitors the electromagnetic environment change data of semiconductor equipment in different process stages.

4. The semiconductor equipment intelligent electromagnetic interference suppression system according to claim 1, characterized in that: The central data processing module includes a data preprocessing unit, a machine learning algorithm unit and an adaptive optimization algorithm unit: The data preprocessing unit is used to receive electromagnetic data from the sensor network and perform preprocessing and feature extraction; The machine learning algorithm unit is used to dynamically model the interference source positioning model based on the machine learning algorithm and to identify and locate the interference source; The adaptive optimization algorithm unit performs behavior prediction analysis on the interference source, generates an interference suppression strategy based on the intelligent control algorithm, and sends an interference suppression instruction to the feedback control module.

5. The semiconductor equipment intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The data preprocessing unit performs multi-level classification on the electromagnetic data received from the sensor network, and extracts high-frequency features, process timing features, and power matching features from the electromagnetic data.

6. The semiconductor device intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The data preprocessing unit performs feature extraction through the following steps: Signal preprocessing step, removing noise and interference clutter from the received original electromagnetic signal; A spectrum feature extraction step, extracting the unique frequency components of semiconductor devices based on a spectrum analysis method; A time domain feature extraction step, extracting time-varying characteristic parameters of electromagnetic signals of semiconductor devices within a specific time period based on a time domain analysis method; The spatial distribution feature extraction step extracts the electromagnetic field distribution features of different regions of the semiconductor device through spatial data fusion of a multi-sensor network; The interference behavior pattern extraction step extracts the behavior pattern of a specific interference source in the semiconductor device.

7. The semiconductor device intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The machine learning algorithm unit analyzes specific interference sources and updates the interference source positioning model in real time as the high-frequency electromagnetic field in the semiconductor equipment changes with the process status.

8. The semiconductor equipment intelligent electromagnetic interference suppression system according to claim 7, characterized in that: The specific interference sources include plasma electrode interference sources and power noise interference sources.

9. The semiconductor device intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The machine learning algorithm unit identifies and locates interference sources through the following steps: Data labeling step: labeling the signal strength and phase information of the preprocessed data and synchronizing the multi-sensor data; Model training and establishment steps: using supervised learning algorithm to train and establish interference source location model, and perform interference source identification; In the interference source positioning step, the phase difference and signal strength difference data are input into the interference source positioning model for calculation to realize the positioning of the interference source.

10. The semiconductor device intelligent electromagnetic interference suppression system according to claim 9, characterized in that: The supervised learning algorithms include support vector machine algorithm, neural network algorithm, random forest algorithm and gradient boosting tree algorithm.

11. The semiconductor device intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The adaptive optimization algorithm unit performs time series analysis and prediction, process specific interference prediction and equipment adaptive capability prediction based on the intelligent control algorithm.

12. The semiconductor equipment intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The adaptive optimization algorithm unit uses a regression model to perform time series analysis and prediction on interference sources.

13. The semiconductor device intelligent electromagnetic interference suppression system according to claim 12, characterized in that: The regression model includes an autoregressive integrated moving average model and a long short-term memory network model.

14. The semiconductor device intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The adaptive optimization algorithm unit identifies the behavior pattern of the interference source by analyzing the interference signal characteristics in the historical data and the real-time data, and predicts the future performance of the interference source according to different operating conditions.

15. The semiconductor device intelligent electromagnetic interference suppression system according to claim 14, characterized in that: The behavior patterns of the interference source include periodic behavior, sudden behavior and random behavior; The adaptive optimization algorithm unit further recognizes the behavior pattern of the interference source, including: The feature vectors of interference sources at different process stages are extracted, and the behavioral patterns of interference sources of semiconductor equipment are identified using a supervised learning model.

16. The semiconductor equipment intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The intelligent control algorithm of the adaptive optimization algorithm unit uses an adaptive interference suppression algorithm and a multi-level feedback control algorithm to generate an interference suppression strategy: The adaptive interference suppression algorithm dynamically adjusts interference suppression parameters according to real-time data of different process stages, combined with the behavior pattern of the interference source and the historical data model; The multi-level feedback control algorithm controls various subsystems of the semiconductor device in real time through a multi-level feedback loop.

17. The semiconductor device intelligent electromagnetic interference suppression system according to claim 4, characterized in that: The intelligent control algorithm of the adaptive optimization algorithm unit also includes dynamic adjustment after interference source identification, intelligent filter self-adjustment mechanism, and predictive control and pre-adjustment.

18. The semiconductor device intelligent electromagnetic interference suppression system according to claim 1, characterized in that: The feedback control module includes an intelligent control unit, an electromagnetic shielding control unit, an adaptive filter unit and a multi-level suppression structure: The intelligent control unit controls parameter variables of the semiconductor device in real time based on control instructions received from the central data processing module; The electromagnetic shielding control unit dynamically adjusts the electromagnetic shielding material based on the control instructions received from the central data processing module; The adaptive filter unit adjusts filter parameters of the semiconductor device based on control instructions received from the central data processing module; The multi-level suppression structure combines multiple feedback loops of the semiconductor device to form a closed-loop control structure with step-by-step feedback, thereby achieving interference suppression of the semiconductor device under different process conditions.

19. The semiconductor device intelligent electromagnetic interference suppression system according to claim 18, characterized in that: The multi-level suppression structure performs multi-level collaborative interference suppression on the interference source layer, the transmission path layer and the receiving end layer.

Images