Radio frequency power supply impedance matching method based on capacitor array

Abstract

The invention belongs to the technical field of radio frequency power supplies, and provides a radio frequency power supply impedance matching method based on a capacitor array, the capacitor array comprises a plurality of capacitor units which are connected in parallel, each capacitor unit is connected with a radio frequency power switch tube, and the radio frequency power switch tubes control the access of the capacitor units; the method comprises the following steps: monitoring a reflection coefficient gamma in real time; if the reflection coefficient gamma exceeds a set range, the pre-trained matching model predicts the plasma concentration and the cavity load impedance at the t + 1 moment based on the plasma concentration change rate and the plasma concentration at the current t moment, and generates a switch tube driving signal to act on a radio frequency power switch tube to control on-off; and inputting the actual plasma concentration and the actual cavity load impedance at the t + 1 moment into the input end of the matching model for adaptive learning. The plasma concentration and the cavity load impedance at the t + 1 moment are predicted, then adjustment of the capacitance value is judged in advance, and the continuous mismatching state between the radio frequency power supply and the plasma cavity is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of radio frequency power supply, and in particular to a radio frequency power supply impedance matching method based on a capacitor array. Background Art

[0002] An RF power matcher, also known as an RF power matcher or an RF load matcher, is a device specifically used to match the impedance between an RF power source and a load. Its principle is based on the basic concept of impedance matching.

[0003] The matcher adjusts the internal electrical components (including inductors, capacitors, transformers, etc.) to change the impedance matching between the power supply and the load, thereby achieving the best energy transmission effect. When the impedance of the RF power supply and the cavity load impedance reach a matching state, the RF power supply can efficiently transfer energy to the load and achieve maximum power transmission, which not only improves the system efficiency, but also effectively reduces the reflection loss of the input signal.

[0004] The RF power matcher is mainly composed of the following core parts: First, inductance. It plays a key role in adjusting the impedance matching between the power supply and the load. It can be in the form of a coil, which provides the required inductance value for the system through its own characteristics, thus helping to achieve impedance matching.

[0005] Second, capacitors. Capacitors are also important components for adjusting the impedance matching between power supply and load. There are two types of capacitors: fixed capacitors and variable capacitors. They can provide corresponding capacitance values ​​according to actual needs and participate in the impedance matching process.

[0006] Third, transformers. In certain cases, transformers can be used to achieve impedance matching. They can change the impedance conversion ratio between the power supply and the load, creating conditions for achieving ideal impedance matching.

[0007] However, given the large output power of the RF power supply, this requires the matcher to have a large power capacity, and the capacitor inside the matcher must be able to withstand higher voltages. In this case, a larger vacuum capacitor is usually selected. Although this capacitor successfully solves the problem of high withstand voltage, it brings a series of new problems. On the one hand, it significantly increases the size and weight of the matcher, and the cost also rises. On the other hand, after using the vacuum capacitor, the matcher can only rely on manual impedance adjustment. This adjustment method is slow and has poor accuracy. When faced with a scenario where the load impedance is in a dynamic change, it cannot meet the needs of fast and accurate adjustment. In addition, this vacuum capacitor-based matcher has limited adjustable bandwidth and frequency range, which further severely limits its scope of use, resulting in its inability to play an effective role in many complex RF application scenarios.

[0008] The RF power matcher plays a key role in impedance matching between the RF power supply and the plasma cavity. As a component of the cavity, the fluctuation of plasma concentration will directly affect the cavity and quickly cause its impedance to change. If the response speed of the matcher cannot keep up with this rapid change, the RF power supply and the plasma cavity will continue to be in a mismatched state, which will cause partial reflection of the RF power output power, reduce the actual power received by the cavity, and have an adverse effect on subsequent processes. Summary of the invention

[0009] In view of the defects in the prior art, the present invention provides an RF power impedance matching method based on a capacitor array to solve the problem that the response speed of the current RF power matcher cannot keep up with the impedance change in the cavity, resulting in a continuous mismatch between the RF power supply and the plasma cavity, which in turn causes partial reflection of the RF power output power.

[0010] The present invention provides a method for impedance matching of a radio frequency power supply based on a capacitor array, wherein the capacitor array includes a plurality of capacitor units connected in parallel, each of the capacitor units is connected to a radio frequency power switch tube, and the radio frequency power switch tube controls the access of the capacitor unit; the method includes: Real-time monitoring of reflection coefficient Γ; If the reflection coefficient Γ exceeds the set range, the pre-trained matching model predicts the plasma concentration and cavity load impedance at time t+1 based on the plasma concentration change rate and the current plasma concentration at time t, and generates a switch tube drive signal to act on the RF power switch tube to control on and off; The actual plasma concentration and the actual cavity load impedance at time t+1 are input into the input end of the matching model for adaptive learning.

[0011] It can be seen from the above technical scheme that the RF power impedance matching method based on the capacitor array provided by the present invention monitors the reflection coefficient in real time. Once the impedance is mismatched, the plasma concentration and the cavity load impedance at time t+1 are predicted by the matching model according to the plasma concentration at the current time t. It can be judged in advance which RF power switching tubes need to be turned on or off, and whether the capacitance needs to be increased or decreased, so as to make full preparations for impedance matching; and then the action of the matcher will not lag, and the reflection of the output power of the RF power supply can be effectively avoided, thereby ensuring a good match between the RF power supply and the plasma cavity, and avoiding the problems caused by the continuous mismatch between the RF power supply and the plasma cavity.

[0012] Optionally, the capacitor unit includes at least one capacitor; when the number of capacitors in the capacitor unit is greater than 1, the capacitors in the capacitor unit are arranged in parallel.

[0013] It can be seen from the above technical solution that there is no need to maintain a one-to-one correspondence between the RF power switch tube and the capacitor; an RF power switch tube can flexibly control one or more capacitors. This design not only improves the matching efficiency, but also effectively broadens the matching bandwidth range and enhances the performance of the matcher.

[0014] Optionally, there is at least one capacitor array; when the number of the capacitor arrays is greater than 1, the capacitor arrays are installed side by side, vertically stacked, or mixedly installed side by side and stacked.

[0015] It can be seen from the above technical solution that more capacitor arrays can be connected in parallel to adapt to tuning and matching networks with different bandwidths and frequencies, providing more possibilities for a wider range of applications. When multiple capacitor array modules are used, side-by-side installation, stacked installation, or mixed installation can be flexibly selected according to the internal space layout of the matcher; based on this, the spatial adaptability of the matcher is enhanced, making it easier to integrate in devices of different sizes and structures.

[0016] Optionally, the capacitor in the capacitor unit is a high-Q chip capacitor.

[0017] It can be seen from the above technical solution that a capacitor array is formed by using chip capacitors and RF power switching tubes to replace traditional vacuum capacitors; the combination of two low-cost devices can greatly reduce the physical size and weight of the capacitor while maintaining excellent electrical performance, directly replacing traditional vacuum capacitors, thereby reducing the volume and weight of the matcher at the source and reducing costs.

[0018] Optionally, the pre-training method of the matching model includes: Inputting a training set into an artificial neural network for training; the training set includes a correlation data set between plasma concentration and cavity load impedance in different circuit states; When the loss function converges to a preset value or reaches a preset number of training times, the training ends.

[0019] It can be seen from the above technical solution that the artificial neural network is trained based on the correlation data set between the plasma concentration and the cavity load impedance in different circuit states; then in specific applications, the neural network is used to predict the future plasma concentration in the cavity in advance based on the current plasma concentration and its rate of change. Since the change in plasma concentration will change the conductivity and electric field distribution of the cavity, thereby affecting the load impedance of the cavity, the change trend of the cavity load impedance can be indirectly predicted through these relationships.

[0020] Optionally, the switch tube driving signal is in differential form; when the signal enters the front stage of the RF power switch tube, it is converted into a single-ended signal through a differential operation circuit.

[0021] As can be seen from the above technical solution, in the internal environment of the matcher, high-power radio frequency signals may cause strong interference to DC signals, thereby affecting the accuracy of controlling the operation of the switch. In this solution, the radio frequency drive signal generated by the matching model adopts a differential form. When this signal enters the front stage of the radio frequency power switch tube, it is converted into a single-ended signal through a differential operation circuit; this process significantly enhances the anti-interference performance of the drive signal and greatly improves the accuracy of the power switch tube in turn-on and turn-off operations.

[0022] Adopting the above technical solution, the present application has the following beneficial effects: The radio frequency power supply impedance matching method based on a capacitor array provided by the present invention can monitor the reflection coefficient in real time. Once the reflection coefficient reflects impedance mismatch, the matching model can predict the plasma concentration and the cavity load impedance at time t + 1 based on the plasma concentration at the current time t, and can determine in advance which radio frequency power switch tubes need to be turned on or off, and whether the capacitance value needs to be increased or decreased, so as to make full preparations for impedance matching; furthermore, the operation of the matcher will not lag, effectively avoiding the reflection of the output power of the radio frequency power supply, ensuring a good match between the radio frequency power supply and the plasma cavity, and avoiding the problems caused by the continuous mismatch between the radio frequency power supply and the plasma cavity. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0024] Figure 1 Shows a flowchart of a radio frequency power supply impedance matching method based on a capacitor array provided by an embodiment of the present invention; Figure 2 Shows a schematic connection diagram of a radio frequency power switch tube and a capacitor unit provided by an embodiment of the present invention; Figure 3 Shows another schematic connection diagram of a radio frequency power switch tube and a capacitor unit provided by an embodiment of the present invention; Figure 4 Shows a flowchart of the operation of a matcher provided by an embodiment of the present invention; Figure 5 Shows a flowchart of the operation of a matching model provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention. It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the invention belongs.

[0026] The RF power matcher plays a key role in impedance matching between the RF power supply and the plasma cavity. As a component of the cavity, the fluctuation of plasma concentration will directly affect the cavity and quickly cause its impedance to change. If the response speed of the matcher cannot keep up with this rapid change, the RF power supply and the plasma cavity will continue to be in a mismatched state, which will cause partial reflection of the RF power output power, reduce the actual power received by the cavity, and have an adverse effect on subsequent processes.

[0027] To solve this problem, Figure 1 As shown, this embodiment provides an automatic impedance matching method based on a capacitor array, comprising: S1. Real-time monitoring of the reflection coefficient Γ.

[0028] S2. If the reflection coefficient Γ exceeds the set range, the pre-trained matching model predicts the plasma concentration and cavity load impedance at time t+1 based on the plasma concentration change rate and the current plasma concentration at time t, and generates a switch tube drive signal to act on the RF power switch tube to control on and off.

[0029] Reflection coefficient If it exceeds the preset range, it means that the current matcher is not in a matching state. At this time, the capacitor needs to be adjusted to achieve impedance matching. is the output impedance of the RF power supply, is the characteristic impedance of the cavity load.

[0030] Specifically, the plasma concentration change rate can be understood as follows: the plasma concentration in the cavity will gradually decrease with the transmission power of the RF power supply, and the speed of concentration reduction is the concentration change rate. The plasma concentration change from t to t+1 is predicted based on the plasma concentration change from t-1 to t (the concentration changes in the two time periods should be consistent or close), and then the plasma concentration at t+1 is indirectly predicted.

[0031] Based on the prediction of plasma concentration and cavity load impedance at time t+1 in this step, it is possible to determine in advance which power switches need to be turned on or off, and whether the capacitance needs to be increased or decreased, so as to make full preparations for impedance matching. In this way, the action of the matcher will not lag, and the reflection of the output power of the RF power supply can be effectively avoided, ensuring a good match between the RF power supply and the plasma cavity.

[0032] Under normal circumstances, the calculation formula of characteristic impedance can be simplified as follows: In this embodiment, the value of the capacitor C is determined by the sum of the capacitors connected in parallel to all the RF power switch tubes in the on state. Assuming that there are k capacitors connected in parallel to the circuit, the total parallel capacitance is .

[0033] If the capacitance value needs to be increased, an RF drive signal is sent to the power switch tube that controls the corresponding capacitor to turn it on, thereby connecting the capacitor in parallel to the circuit and increasing the total capacitance value of the circuit; conversely, if the capacitance value needs to be reduced, the RF drive signal is stopped from being sent to the power switch tube that controls the capacitor to turn it off, thereby disconnecting the capacitor from the circuit and reducing the total capacitance value of the circuit.

[0034] S3. Input the actual plasma concentration and the actual cavity load impedance at time t+1 into the input end of the matching model.

[0035] In this step, the actual plasma concentration and impedance matching effect after the switch tube is driven are fed back to the input end of the neural network to continuously correct the prediction deviation. Each feedback is an "online learning" that allows the neural network to dynamically track scene changes and maintain the accuracy of the output drive signal; it enables the neural network to have adaptive learning capabilities and automatically adjust its internal parameters and structure based on new data and experience to adapt to changing loads and demands. Optimization algorithms in artificial intelligence technology (such as genetic algorithms, particle swarm optimization, etc.) can also be used to further adjust the parameters or structure of the neural network to improve its calculation of impedance matching and real-time control of the performance of the switch tube.

[0036] It should be noted that in this embodiment, an RF power switch tube is used to control the on and off of the capacitor unit. Compared with the ordinary switch method, the RF power switch tube is highly sensitive to frequency and is very suitable for situations where high-speed matching is required. The switching speed of the RF power switch tube is affected by parasitic parameters, and there will be parasitic capacitance and parasitic inductance at high frequencies, which will affect the performance of the matching circuit. Therefore, in this embodiment, an RF power switch tube with low parasitic capacitance and parasitic inductance and a material with high carrier mobility, such as a switch tube made of GaN (gallium nitride) or SiC (silicon carbide) material, can be selected. At the same time, using high-frequency PCB materials, the traces in the matching circuit should be as short as possible and the line width should be uniform to avoid impedance discontinuity and signal reflection. In addition, it is also necessary to avoid crossing and interference between traces to ensure the purity and stability of the signal. At the same time, grounding design is also crucial, and good grounding needs to be ensured to reduce noise and interference.

[0037] Meanwhile, in this embodiment, multiple capacitor arrays can be placed in parallel. Connecting more capacitor arrays in parallel can adapt to tuning and matching networks with different bandwidths and frequencies, thereby expanding the scope of application.

[0038] Through the flexible control of the RF power switch tube, it was unexpectedly discovered that the circuit's adaptability to changes in external conditions has been enhanced. For example, in application scenarios with large load changes, the circuit can respond and adjust its working state more quickly to maintain stable impedance matching performance. Precise control of the RF power switch tube inadvertently broadens the circuit's adaptability to different loads. Appropriate on-off scheduling of the RF power switch tube can make the energy distribution and transmission within the circuit smoother and more orderly. The voltage and current jitter that may have been caused by impedance fluctuations will be effectively buffered, improving the stability of the system.

[0039] Specifically, the switch tube drive signal is in differential form; when the signal enters the front stage of the RF power switch tube, it is converted into a single-ended signal through a differential operation circuit. In the internal environment of the matcher, high-power RF signals may cause strong interference to DC signals, thereby affecting the accuracy of controlling the switch action. The RF drive signal generated by the matching model in this scheme is in differential form. When the signal enters the front stage of the RF power switch tube, it is converted into a single-ended signal through a differential operation circuit; this process significantly enhances the anti-interference performance of the drive signal and greatly improves the accuracy of the power switch tube in the opening and closing operations.

[0040] like Figure 2-3 As shown, the capacitor unit includes at least one capacitor. Figure 3, when the number of capacitors in a capacitor unit is greater than 1, the capacitors in the capacitor unit are set in parallel, and then connected to the RF power switch tube, and the power switch tube determines the on and off of the connected capacitor unit. In the same capacitor array, the number of capacitors set in different capacitor units is not exactly the same. For example, it can be set that there is 1 capacitor in m capacitor units, 2 capacitors in n capacitor units, 3 capacitors in o capacitor units... In this way, when it is necessary to control the access of capacitors with different capacitance values ​​in step S2, a capacitor unit containing an appropriate number of capacitors can be selected to directly control its access, thereby reducing the amount of operation and improving control efficiency.

[0041] in, Figure 2 This is a schematic diagram of a RF power switch tube controlling a capacitor parallel circuit, that is, only one capacitor is set in the capacitor unit. Each RF power switch tube is responsible for regulating a capacitor. The capacitance of these capacitors can be kept consistent or gradually increased or decreased in a certain order, but the key is that the change in capacitance must be maintained within a small step range to ensure the accuracy of impedance matching and avoid jumps during the matching process. The neural network performs matching calculations based on the actual load impedance, determines the required increase or decrease in capacitance, and sends an RF drive signal to the corresponding RF power switch tube accordingly, thereby controlling whether the capacitor regulated by the switch tube is connected to the bus or disconnected from the bus to achieve parallel or disconnection operations. Given that the RF power switch tube is highly sensitive to the frequency of the drive signal, its opening and closing actions are faster than ordinary switches. Combined with the predictive control capability of the neural network, the RF power switch tube exhibits superior performance than ordinary switches in application scenarios that require higher matching speeds.

[0042] in, Figure 3 The schematic diagram of a multi-capacitor parallel circuit controlled by a single RF power switch tube, that is, multiple capacitors are set in a capacitor unit. Generally speaking, a capacitor that can withstand high voltage and has a high Q value has a Q value greater than 1000; its capacity is often small, with a capacitance less than or equal to 1nF. If each capacitor is controlled by a separate RF power switch tube, the space occupied by the printed circuit board will increase, which will pose an obstacle to the miniaturization and integration of the capacitor array module. In order to overcome this problem, multiple capacitors can be reasonably connected in parallel to the source of the power switch tube to increase the total amount of capacitance controlled by each power switch tube, thereby reducing the number of power switches required. This method helps to achieve the miniaturization and integration of capacitor arrays.

[0043] Furthermore, in this embodiment, the capacitor units are all chip capacitors. Based on the chip capacitors and the RF power switch tubes to form a capacitor array, the combination of two low-cost devices can significantly reduce the physical size and weight of the capacitor while maintaining excellent electrical performance, directly replacing the traditional vacuum capacitor, thereby reducing the volume and weight of the matcher at the source and reducing the cost.

[0044] Based on this, in specific implementation, a certain number of chip capacitors can be evenly welded on the upper and lower surfaces of the printed circuit board, and the capacitors on the upper and lower surfaces are connected in parallel through the through holes on the printed circuit board to form a capacitor array to replace the traditional vacuum capacitor.

[0045] The number of capacitor arrays can also be set according to actual needs. There is at least one capacitor array; when the number of capacitor arrays is greater than 1, the capacitor arrays are installed side by side, vertically stacked, or mixed side by side and stacked. The following is a detailed description of different installation methods: 1. In the side-by-side installation mode, each capacitor array module is placed horizontally and arranged side by side in the internal space of the matching device. The advantages of this method are: Easy to dissipate heat: Since there is a certain distance between the modules, the air can circulate smoothly, which is conducive to heat dissipation and avoids overheating.

[0046] Easy maintenance: The independent layout between modules makes it easier to replace or repair a single module without disassembling the entire matcher.

[0047] High space utilization: When the internal space of the matcher allows, the number of modules can be increased to the maximum, thereby improving the performance of the matcher.

[0048] 2. The stacking installation method is to stack the capacitor array modules vertically to save horizontal space inside the matcher. This method is applicable to the following scenarios: Limited space: When the internal space of the matcher is limited, the stacking installation method can more effectively utilize the vertical space and avoid wasting horizontal space.

[0049] High integration: The stacking method enables the matcher to achieve higher integration while maintaining high performance, making it easier to integrate in compact devices.

[0050] Stability requirements: In some applications, in order to maintain the stability of the matcher, it may be necessary to stack the modules closely together to reduce interference from the external environment.

[0051] 3. In addition to the two basic installation methods mentioned above, mixed installation methods are also supported. That is, according to the specific conditions of the internal space of the matcher and the matching requirements, you can flexibly choose a combination of side-by-side and stacking. This mixed installation method can maximize the use of the internal space of the matcher while meeting the performance requirements of different application scenarios.

[0052] This embodiment also provides a pre-training method for a matching model, including: The training set is input into the artificial neural network for training; the training set includes a correlation data set between the plasma concentration and the cavity load impedance in different circuit states; When the loss function converges to a preset value or reaches a preset number of training times, the training ends.

[0053] Artificial neural networks can consider using architectures such as multi-layer perceptron (MLP), convolutional neural network (CNN) or recurrent neural network (RNN). Select appropriate activation functions such as ReLU, Sigmoid, etc. to enhance the nonlinear expression ability of the neural network. And set appropriate loss functions such as mean square error (MSE) or cross entropy loss to measure the difference between the RF power switch control signal output by the neural network and the actual expected output.

[0054] When training a neural network, a large amount of training data needs to be collected, including input parameters and corresponding correct switch control signals under different circuit states. The neural network is then trained based on the training data; optimization algorithms such as gradient descent can be used to adjust the weights and biases of the neural network to minimize the loss function. Finally, multiple iterations of training are performed until the loss function converges to a smaller value, or the preset number of training times is reached, so that the output RF power switch tube control signal is infinitely close to the actual expected output.

[0055] The implementation of the artificial neural network provided in this embodiment is as follows Figure 5 As shown. First, the input signal, state information and other data must be collected and preprocessed, which is the data that the artificial neural network needs to input and judge. Then build a neural network model, including construction, determining input and output, selecting activation functions and setting loss functions. After the artificial neural network model is built, a large amount of data can be manually given to train the network and iterate multiple times to achieve accurate analysis and calculation of impedance by the artificial neural network. After training, the artificial neural network can predict the impedance matching of the circuit and formulate a control algorithm to adjust the state of the power switch tube. The neural network outputs the control signal of each power switch tube to determine which switches should be turned on or off to achieve a specific capacitor array configuration. After the decision is made, the decision is executed, a control signal is sent to the power switch tube to change its state, and a feedback mechanism is implemented to monitor the control effect. If the reflection coefficient of the matcher does not meet expectations, the algorithm can adjust the control algorithm or retrain the model to improve accuracy. Finally, using intelligent optimization and adaptive learning technology, its internal parameters and structure are automatically adjusted to optimize the matching degree between its calculation source and load and the performance of real-time control of the switch tube.

[0056] Based on this, the matching device is based on a RF power impedance matching method based on a capacitor array provided in this embodiment, and the working process is as follows: Figure 4As shown. First, several capacitor arrays are placed in parallel inside the matcher to facilitate subsequent impedance matching; when the matcher enters the working state, the matching model dynamically scans the impedance and analyzes the impedance matching state between the source and the load. According to the degree of mismatch, the matching model quickly and accurately controls the on and off state of the RF power switch tube, and then selectively connects the capacitors in the capacitor array in parallel to the matching circuit, thereby achieving precise impedance matching, which is the key to achieving efficient energy transmission and reducing energy loss. In some application scenarios, such as advanced plasma etching processes, the load is not static, which requires the matching model to instantly adjust the switching state of the RF power switch tube according to the real-time changes of the load, and accurately control the number of capacitors connected in parallel to the matching circuit. This real-time, dynamic adjustment mechanism ensures that the source and the load can achieve a matching state regardless of how the load changes.

[0057] Based on the method provided in this embodiment, in specific application, the following steps are included: In the process of constructing the capacitor array, the frequency range of the preset matcher is [F1, F2], and the bandwidth adjustment range is set to [B1, B2]. After calculation and analysis, we selected a set of capacitance values ​​C1, C2, ..., Cn to form the capacitor array. Each capacitor unit Ci (where i = 1, 2, ..., n) is connected to a power switch tube Si. When Si is in the on state, the corresponding capacitor unit Ci will be connected in parallel to the circuit; otherwise, the capacitor unit Ci will be disconnected from the circuit.

[0058] First, the matching model is implemented based on an artificial neural network, which aims to enable the neural network to learn to adjust the capacitance value and generate the on-drive signal of the power switch tube accordingly according to the impedance matching requirements in various situations, and calculate the reflection coefficient to determine whether the reflection coefficient exceeds a given range.

[0059] Next, we provided the neural network with a data set between plasma concentration and cavity load impedance, allowing it to quickly identify the relationship between the two and adjust the drive signal of the matching capacitor accordingly. We then trained it to directly control the conduction state of the power switch tube according to the plasma concentration to achieve impedance matching.

[0060] Ultimately, the neural network needs to learn the rate of change of plasma concentration and predict changes in cavity load impedance based on this trend, so that it can pre-adjust the drive signal of the power switch tube and change the matching condition of the circuit in time.

[0061] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0062] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for impedance matching of a radio frequency power supply based on a capacitor array, characterized in that: The capacitor array includes a plurality of capacitor units connected in parallel, each of which is connected to a radio frequency power switch tube, and the radio frequency power switch tube controls the access of the capacitor unit; the method includes: Real-time monitoring of reflection coefficient Γ; If the reflection coefficient Γ exceeds the set range, the pre-trained matching model predicts the plasma concentration and cavity load impedance at time t+1 based on the plasma concentration change rate and the current plasma concentration at time t, and generates a switch tube drive signal to act on the RF power switch tube to control on and off; The actual plasma concentration and the actual cavity load impedance at time t+1 are input into the input end of the matching model for adaptive learning.

2. The method according to claim 1, characterized in that: The capacitor unit includes at least one capacitor; when the number of capacitors in the capacitor unit is greater than 1, the capacitors in the capacitor unit are arranged in parallel.

3. The method according to claim 2, characterized in that There is at least one capacitor array; when the number of the capacitor arrays is greater than 1, the capacitor arrays are installed side by side, vertically stacked, or mixedly installed side by side and stacked.

4. The method according to claim 3, characterized in that The capacitor in the capacitor unit is a high-Q chip capacitor, and the Q value is greater than 1000.

5. The method according to any one of claims 1 to 4, characterized in that: The pre-training method of the matching model includes: Inputting a training set into an artificial neural network for training; the training set includes a correlation data set between plasma concentration and cavity load impedance in different circuit states; When the loss function converges to a preset value or reaches a preset number of training times, the training ends.

6. The method according to claim 5, characterized in that The switch tube driving signal is in differential form; when the signal enters the front stage of the radio frequency power switch tube, it is converted into a single-ended signal through a differential operation circuit.

Images