Method, system and data carrier signal for machine learning in plasma systems - Patents.com

Abstract

The present invention relates to a method for machine learning detection of at least one anomaly in a plasma system, in particular in an RF powered plasma processing system, comprising: providing at least one input signal (210) having at least one anomaly characteristic indicative of an anomaly in the plasma system, said input signal (210) being related to an analog signal of a power supply system (1) of the plasma system and / or another characteristic of the power supply system (1) and / or the plasma system; The method includes performing a machine learning procedure (310) and processing at least one input signal (210) having at least one anomaly feature by a programmable circuit (10) to train detection of anomalies in a plasma system.

Description

[Technical Field]

[0001] The present invention relates to a method for machine learning detection of at least one anomaly in a plasma system, and to a system and a data carrier signal. [Background technology]

[0002] Digital signal processing, or "DSP," can be used in radio frequency systems for a variety of applications, such as filtering and / or detecting specific system conditions. To improve processing speed and / or performance, DSP can be implemented at least in part in hardware, for example, by an integrated circuit, or "IC." However, parameterizing an integrated circuit can be time-consuming and technically complex. The complexity of the parameterization process can affect digital signal processing in many undesirable ways. One of these is a partial reduction in the DSP efficiency of the integrated circuit.

[0003] Plasma power supply systems can measure radio frequency (RF) signals. Typical RF measurements include forward and reverse power, as well as RF current and voltage at various points within the plasma power supply system. Measured signals typically vary in frequency, amplitude, noise or spurious content, and pulse length. While some variations, such as frequency, may be well controlled by the power supply system, other variations may not be controlled, especially when delivering RF power to a plasma processing chamber. These variations can introduce various types of interference into the measured RF signal. For example, one form of interference can be harmonics generated by the nonlinear nature of the plasma itself. Another form of interference can result from different frequencies used to generate the RF signal. In advanced plasma systems, several RF generators with different frequencies may be used to excite the plasma. However, in addition to the desired generation of the intended set of frequencies, mixed frequencies, such as combinations of a fundamental frequency and its harmonics, may also be generated.

[0004] Many interferences found in measured RF signals can be suppressed or at least reduced by using appropriate filters before digitizing the measured RF signal. Nevertheless, unwanted sidebands close to the base frequency may occur, necessitating or imposing additional filtering requirements on the DSP. Real input signals are never ideal because they always contain unavoidable distortions and other artifacts caused by the signal source, transmission line, measurement equipment, etc. Optimizing the filter or filter parameters is crucial to reconstruct an ideal, distortion-free input signal for stable network regulation. However, each filter has advantageous and disadvantageous characteristics. For example, if filtering increases signal delay too much, the control will not be able to respond in a timely manner to deviations from the intended power level. Furthermore, filtering can remove relevant portions of the signal that may be needed to evaluate the quality of the plasma process, such as the presence of anomalies like secondary plasma (which may exist outside the intended area of ​​the plasma chamber), arcs, and other plasma instabilities. Therefore, balancing these characteristics to optimize performance requires careful selection of not only the filter type but also its parameter configuration.

[0005] In general, insufficient interference reduction can cause fluctuations in measurements of the output power of an RF generator. These fluctuations can be understood as "measurement artifacts." Because measurements may be used in various applications, such as controlling output power levels, these applications may rely on a compromised measurement signal. In the control example, the output power level may erroneously follow the measurement artifacts rather than the correct measurement.

[0006] If plasma instabilities are suppressed by filtering, anomalies such as secondary plasma or arcing may not be detected, and as a result, no corrective action may be initiated. In semiconductor plasma processing systems, plasma instabilities such as uncontrolled arcing can damage the semiconductor wafers being processed.

[0007] US20060262889A1 discloses a power supply that can generate a signal when driven by a drive signal from a controller. At radio frequencies, either synchronous undersampling or synchronous oversampling applies.

[0008] In DE102015212242A1 / US20180122625A1, the sampling frequency is changed to remove the interference. Both documents describe specific methods for compensating for the interference frequency, including adapting the sampling frequency, but do not describe machine learning, in particular the optimization of filter parameters by machine learning procedures.

[0009] The general use of programmable circuits, in particular field programmable gate arrays (FPGAs), to accelerate simulations may be known from the prior art, for example from WO2007095574A1 or DE102011103861A1.

[0010] Furthermore, it is known from the prior art that intelligent systems exist for detecting certain situations during plasma processing, however, this detection often lacks the reliability and speed to initiate corrective action before the event is completed in time.

[0011] The use of dedicated solutions for detecting specific situations in plasma processing has been reported in the prior art, for example in US8989888B2.

[0012] In particular, the use of dedicated solutions to manage arcing in plasma processing has been reported in several prior art documents.

[0013] EP 3254295 B1 describes, inter alia, an arc treatment device and method for suppressing arcs by turning off RF power for specific periods of time.

[0014] US9316675B2 describes a method for detecting secondary plasma by monitoring changes in RF voltage and current, which can distinguish between low-level anomalies and severe anomalies.

[0015] In US8890537B2, arcs can be detected by appropriate filtering operations and analysis of harmonics.

[0016] The use of dynamic limiting of RF related signals such as reflected power, voltage, current, etc. to identify arcs, as well as various methods for suppressing arcs, are described, inter alia, in US Pat. No. 7,761,247 B2.

[0017] US10209294B2 shows how the detection of an arc can be improved by combining several observation signals. Summary of the Invention [Problem to be solved by the invention]

[0018] These prior art documents describe hardware or software solutions tailored to operate under specific process conditions, such as in sputtering systems, or with specific anomalies (e.g., secondary plasma or arc only). Notably, all solutions require an actual anomaly to be present before the system detects it. In the case of arcs, the described arc management methods specifically only work to suppress further arcs after at least one arc event has already occurred.

[0019] In contrast, the invention described below may have wider applicability and be more versatile due to the fact that it uses a statistical approach to detect abnormalities before they fully develop and, in particular, can prevent their occurrence.

[0020] It is therefore an object of the present invention to at least partially overcome the above-mentioned drawbacks, and in particular to improve digital signal processing in RF systems.

[0021] The above-mentioned object is solved by a method having the features of claim 1, a system having the features of claim 16, and a data carrier signal having the features of claim 17. Further features and details of the invention are disclosed in the respective dependent claims, the description and the drawings. Features and details explained in connection with the method according to the invention also apply to the system according to the invention and to the data carrier signal according to the invention, and vice versa in each case. The individual aspects of the invention can always refer to each other. [Means for solving the problem]

[0022] In contrast to previous documents in the prior art, the method, system, and carrier signal of the present invention make it possible to optimize digital signal processing, particularly in radio frequency systems. For this purpose, it is particularly an idea of ​​the present invention that a processing procedure can be used, particularly for training the inventive system. Training can be performed using reference data in a training mode to determine optimized parameterizations of the programmable circuit and / or algorithm, after which the inventive system can be switched to an application mode in which the optimized parameterizations are used in real time to process the relevant input signals. The programmable circuit can be a programmable signal processing circuit, such as an FPGA, for processing input signals by executing algorithms.

[0023] Further advantages of the method, system, and carrier signal of the present invention over the prior art may be achieved by performing training and real-time operation modes in the same system. Furthermore, further training using a new training data set, i.e., machine learning of a new situation, can be advantageously performed at a later time, and the existing parameterization can be further optimized (hereinafter, this is specifically referred to as "switching processing from real-time mode to training mode"). Furthermore, it is also possible to transfer the optimized parameterization to other systems of the same type, for example, by using a user interface.

[0024] Advantageously, using the same hardware components (FPGA, control, memory) in both training and application modes facilitates the transfer of parameterization between systems, enabling future upgrades of plasma processing systems for new, as-yet-unknown situations. Training can be performed on a test bench system or laboratory plasma processing system, for example, using existing input signals recorded on a system used in the field. The variability of these signals can be numerically increased by scaling, filtering, superposition, etc. The optimized parameterization can then be easily transferred from the training system to the field plasma processing system.

[0025] Additionally, by using existing input signals from memory, the functionality of an RF system can be tested in a variety of situations, for example, during factory acceptance testing of the RF system, where input signals are provided by memory in place of or in combination with ADC data, and system operations such as regulation loop performance, RF signal filtering, pattern recognition, and response actions, e.g., alarms or interlocks, can be verified with a fully operational system.

[0026] The object is solved in particular by a method for machine learning detection of at least one anomaly in a plasma system, in particular a radio frequency (short RF) power plasma system, preferably an RF power plasma processing system, which may comprise (performed one after the other or in any order): providing at least one (or more) input signals, each of which relates to an analog signal of a power supply system of the plasma system and / or another characteristic of the power supply system and / or the plasma system and has at least one anomaly characteristic indicative of an anomaly in the plasma system; · Implementing a machine learning procedure, wherein at least one input signal having at least one anomaly feature is processed by a (preferably programmable) circuit to train the detection of anomalies in the plasma system, in particular this processing being at least partly performed by the detection procedure.

[0027] This has the advantage of providing effective and fast machine learning. In particular, the use of (preferably programmable) circuits for the machine learning procedure can further accelerate this process. Furthermore, the detection of anomalies can provide safer and more reliable operation in plasma systems. According to another aspect of the present invention, a machine learning procedure can be configured to be at least part of a processing procedure. When performing the processing procedure and / or the machine learning procedure, at least one input signal can be repeatedly processed by a circuit, preferably a programmable circuit, preferably a detection procedure, using at least one configurable parameter of the process (i.e., a process performed in particular within the processing procedure and / or the machine learning procedure, preferably in the form of digital signal processing). The at least one configurable parameter can be at least one parameter of the detection procedure. The configuration of the at least one parameter can be changed, in particular modified, for each processing of at least one input signal within the processing procedure and / or the machine learning procedure to obtain a respective processing result. The configurable parameter can be a parameter of the detection procedure, in particular a parameter of a neural network weight or algorithm. The detection procedure can include digital signal processing in a radio frequency system, in particular a radio frequency power supply system. The respective processing result can be a result or output of the detection procedure. The configuration can be changed to improve the detection procedure and thus the digital signal processing.

[0028] The at least one input signal may relate to a radio frequency signal of the radio frequency system and / or another characteristic of the power supply system and / or plasma system, for example a signal from the plasma chamber (such as an optical signal), preferably a signal measured in the plasma chamber, and / or a signal of an external source.

[0029] The method according to the present invention may further comprise the following. determining at least one parameter result, in particular a modified configuration selection, preferably a set of weights and / or values, for the optimization of the configuration based on the processing results, in particular as a machine learning training result, Providing at least one determined parameter result, in particular a determined configuration, for improved digital signal processing, in particular for optimization of a circuit using the parameter result and in particular the selection.

[0030] The provided and / or determined parameter results, preferably including a weight set for a neural network, may be used to parameterize the circuit. The parameterized circuit may perform a detection procedure, which may include application of a neural network using the parameter results, in particular the weight set provided by the parameter results. Thus, the circuit may implement a neural network, or in other words, a neural network that has been parameterized using the parameter results, in particular the weight set.

[0031] The parameterization of the circuit may include parameterization of the detection procedure according to the parameter results, i.e., preferably weights of the neural network according to the parameter results. The performance of the parameterized detection procedure may be aimed at detecting at least one anomaly in normal (in-situ) operation of the plasma system.

[0032] Optionally, the programmable circuit may be configured as a programmable integrated circuit, preferably a digital signal processor, a complex programmable logic device, or a field programmable gate array.

[0033] Advantageously, within the scope of the present invention, it may be provided that a programmable circuit, in particular, executes a detection procedure at least in part, which includes pattern recognition or pattern matching, in particular using neural applications, preferably neural networks, and / or algorithms, for identifying at least one anomalous feature. The weights and / or activation functions and / or neurons of the neural network may be provided by the detection procedure implemented in particular by the circuit. Some of the activation functions and / or weights may be manually predefined.

[0034] The machine learning procedure may include: - the processing of at least one input signal is carried out by the circuit using the detection procedure and at least one configurable parameter of the detection procedure, in particular a weight of a neural network or a parameter of an algorithm, and the configuration of the at least one parameter is changed, in particular modified, for each processing of the input signal to obtain a respective processing result, Determining at least one parameter result, including in particular the selection of a modified configuration, as a machine learning training result based on the processing result.

[0035] The parameter results and / or training results may be referred to as the classification of the neural network.

[0036] Furthermore, the parameter results may be distributed for use in other circuits implementing the detection procedure. A data carrier signal, preferably a data carrier having a data carrier signal, may contain the parameter results for this distribution. Therefore, a data carrier signal or a data carrier containing the parameter results is also an object of the present invention.

[0037] Within the scope of the present invention, machine learning procedures, particularly processing of at least one input signal, include repeated processing steps in which the same at least one input signal is processed, particularly by a programmable circuit, but using modified configurations with different parameters to obtain respective processing results, each assigned to the configuration used. Preferably, evaluation of the modified configurations is performed by comparing each obtained processing result with a reference result. The evaluation may be part of a machine learning procedure that allows training in the sense of optimizing a configuration (e.g., by using a conventional optimization algorithm). Thus, machine learning is not only understood within the scope of artificial intelligence, but may also, additionally or alternatively, involve the automatic solution of an optimization problem by an optimization algorithm, or simply the repeated search for the best parameters of an algorithm to identify at least one anomalous feature. The evaluation may be performed by another device external to the control circuit or the programmable circuit.

[0038] Also, optionally, it may be possible that the evaluation of the modified configuration depends on matching the processing results with the reference results, and at least one configuration having the highest evaluation is selected from the modified configurations as a result of the at least one determined parameter.

[0039] Additionally, the reference result is optionally provided to be a predetermined indicator of at least one abnormality, and thus depends on how close the processing result is to the ideal reference result.

[0040] According to an advantageous further development of the invention, it can be provided that each of the at least one input signal is related to: Radio frequency signals used to power the plasma system, and / or Another characteristic of the power supply system and / or plasma system, Preferably, the anomaly is specific to arcing occurring in the plasma processing system or the probability of arcing occurring. The anomaly may be a particular pattern indicative of a particularly expected arc or other undesirable anomaly occurrence.

[0041] Arcing can be understood as an electrical flashover or arc discharge, especially during plasma processing. Particularly in plasma-assisted coating processes, this anomaly can disrupt the plasma coating. It is therefore advantageous if the detection procedure can detect arcs in the shortest possible time.

[0042] Optionally, the machine-learned detection of at least one anomaly may be used for arc detection and / or arc prevention and / or arc management. Detection may result in a warning message being output. Further action, such as shutoff, enables arc management. If a possible arc is detected, arc prevention is enabled.

[0043] Preferably, a machine learning procedure may be provided that provides for the iterative determination of the configuration of at least one parameter of the detection procedure, in particular the configuration of at least one weight of a neural network or at least one configuration of an algorithm for detection. In particular, the neural network or the algorithm is at least partially implemented by a programmable circuit. The determined configuration is then used in the detection procedure in field operation of the power supply system and may include: - When an abnormality is detected by the detection procedure, warning information is output.

[0044] Also advantageously, providing at least one input signal comprises: recording and converting analog signals of the power supply system in the form of radio frequency signals and / or other characteristics of the power supply system and / or plasma system to obtain a digital representation of the analog signals and / or other characteristics; · Providing an analog signal and / or a digital representation of another characteristic as at least one input signal, particularly to a programmable circuit.

[0045] As a further advantage, it may be provided that the detection procedure for detecting at least one anomaly is implemented in a particularly programmable circuit by combining functional blocks of the circuit for performing calculations in real time, which also allows for a fast execution of the neural network. It is also conceivable within the scope of the present invention that the implemented detection procedure is used repeatedly in real-time calculations during a machine learning procedure for processing at least one input signal, thereby making it possible to accelerate the machine learning.

[0046] It is also conceivable that the machine learning procedure is configured as a training procedure, in particular for the parameterization of a programmable circuit, so that the exact same circuit or another circuit with the same parameterization can be used for the detection of at least one anomaly during on-site operation of the power supply system, in particular for arc detection and / or arc prevention and / or arc management.

[0047] It may also be provided that, particularly in field operations, upon detection of at least one anomaly, at least one of the following actions is initiated, preferably individually or in combination: - the interruption of at least part of the electricity supply system or the complete interruption, in particular the interruption of only part of the electricity supply system or the interruption of all parts of the electricity supply system; - Power supply system shutdown without restarting the power supply system -Temporary shutdown of the power supply system and subsequent restart of the power supply system; -Temporary reduction in the output power of the power supply system; temporary modification of the output frequency of the power supply system, in particular the output frequency of the RF power signal and / or analog signal of the power supply system; ·Temporary correction of at least one adjustable element of the impedance matching network.

[0048] This allows for effective arc management and / or prevention. Generally, a temporary modification of at least one adjustable parameter of the power supply system, e.g., a change in RF power signal frequency or a change in the value of an adjustable component of the impedance matching circuit, may be initiated upon detection of at least one anomaly, preferably individually or in combination with the above-described actions. The exact parameters used for this action may be predefined depending on the specifications and exact configuration of the power supply system. The most suitable parameters for this may be found experimentally or empirically using a particular power supply system.

[0049] It is further advantageous within the scope of the present invention if the detection of anomalies includes probabilistic detection, in particular for detecting arcs before they occur in the plasma processing system, in other words for detecting anticipated arcs, preferably for determining the probability of arc occurrence and / or for arc prevention. The detection of anomalies may also include probabilistic detection for detecting other anomalies before they occur in the plasma processing system.

[0050] Another aspect of the present invention is a system for detecting at least one anomaly in a plasma system, particularly an RF powered plasma system, preferably an RF powered plasma processing system, comprising: at least one sensor, in particular a directional coupler and / or a voltage / current sensor, for detecting at least one radio frequency signal of the plasma supply system of the plasma system and / or another characteristic of the power supply system and / or the plasma system; at least one converter, in particular an analog-to-digital converter, connected to the sensor for converting the at least one detected radio frequency signal and / or another characteristic into at least one input signal; A circuit, in particular a programmable circuit, for digital signal processing of at least one input signal, trained to detect at least one anomaly according to the method according to the invention. The system according to the invention therefore offers the same advantages as those explained in detail with reference to the method according to the invention. The method according to the present invention has the advantage that multiple processing results, each specific to the parameter configuration used, are available, and the best configuration can be selected from the modified configurations based on the processing results. In other words, parameters can be optimized by modifying the configuration (also called "modifying the parameters"). Thus, the selection of the parameter result may include at least one of the modified configurations that provides the best processing result within the processing sequence. The use of the parameter result for parameterizing a circuit can achieve improvements in digital signal processing (DSP).

[0051] The processing procedure may involve iterative processing by a circuit, in particular digital signal processing, but may also include further steps such as evaluating and / or modifying the configuration, in particular as described in more detail below. Furthermore, the processing procedure may be considered at least in part as a machine learning procedure. In other words, the processing procedure may include a machine learning procedure. The machine learning procedure may include (at least) evaluating and / or storing determined parameter results in a memory, and / or determining and / or providing parameter results, and / or learning and / or optimizing a configuration, preferably parameter values ​​and / or at least one weight of a neural network.

[0052] Within the scope of the present disclosure, indefinite and definite articles or numerical designations such as "1", "2", etc., should be understood as designations of "at least", unless expressly stated otherwise. Even when there is an express designation of "at least" or a similar designation, it should not be concluded that a limitation in the sense of, for example, "exactly 1" is implied by another simple use of the article or numerical designation without this express designation. Furthermore, numerical designations, as well as designations of process parameters and / or equipment parameters, should likewise be understood in the technical sense, i.e., with normal tolerances.

[0053] The use of the term "best processing results" means the following: Optimal results of digital signal processing, e.g., maximum suppression of unwanted frequencies or maximum number of specific anomalous signals detected among many anomalous signals. Optimal results for measurement parameters related to the RF power delivery system, e.g. RF power output with least fluctuations. Optimal detection of a specific pattern in one of the measured signals. Optimal results for measuring parameters related to plasma processing, e.g. DC self-bias or ion concentration. Optimal results for measuring parameters related to the properties of the article being treated in the plasma chamber, e.g. the thickness of the deposited layer.

[0054] To provide at least one input signal, the execution of the processing procedure may be initiated and controlled by a control circuit (hereinafter also referred to as "control") electrically connected to the programmable circuit, and the determining and providing may be performed by the control circuit.

[0055] The term "digital signal processing" is generally understood to mean processing of digital signals. Specifically, digital signal processing according to the present invention may include processing of RF signals after they are sensed by at least one sensor and then converted to digital signals in real time by an analog-to-digital converter, or ADC, in the RF system. In contrast to digital signal processing, a "processing procedure" may be considered a training procedure that does not use input signals that are actually sensed in real time (instead, they may be pre-recorded or computer-simulated). The algorithms used for processing in the processing procedure and in the digital signal processing may be the same. However, the algorithm configuration is first trained in the processing procedure and then applied in the digital signal processing.

[0056] The term "RF system" refers to a system that outputs at least one electrical RF signal. The RF system may also include or be part of an RF generator. Thus, the RF signal may be the output of an RF generator. The term radio frequency power supply system may refer to a specific configuration of an RF system, specifically the power supply of a plasma system, particularly a plasma generator system. The system may further include transmission lines and / or impedance matching networks. The plasma system may be suitable for generating plasma used in plasma processing, preferably plasma etching or plasma deposition. As is well known to those skilled in the art, plasma processing applications typically require amplified power signals of 100 W or more, often exceeding 1 kW, to adequately ionize gases and generate plasma in a plasma processing chamber. However, the control signals involved in controlling these high-power signals may involve much lower power signals and may be processed according to low-power signal processing techniques. Sensing the characteristics of high-power signals, such as RF signals, may require various types of sensors and may involve picking up features of the high-power signal, including a small fraction of its amplitude.

[0057] The "at least one input signal" may preferably be a digital signal related to a radio frequency signal (abbreviated as RF signal, also referred to as RF power signal) of a radio frequency system. In particular, this means that the at least one input signal may be specific to the RF signal and / or may be obtained by detecting, in particular measuring, the RF signal. Thus, the RF signal may be a physical signal in the radio frequency range that is the actual electrical output of or within the RF system. The RF range may be a frequency range between 20 kHz and 300 GHz, preferably between 300 kHz and 300 MHz (mid-, high-, and ultra-high-frequency bands). The input signal may also be a low-frequency-varying DC signal derived from the RF signal, e.g., in pulsed applications, the measured DC bias voltage present in the plasma chamber, the measured forward and / or reflected power, the reflection coefficient, the phase between RF voltage and current, or a scalar signal from an external sensor.

[0058] After the machine learning procedure is performed, the trained detection of anomalies can be used to control the signal generating unit. In other words, in field operation, the parameterized programmable circuit can be used to detect anomalies using the detection procedure, and when at least one anomaly is detected, control of the signal generating unit can be initiated. At least one anomaly can be detected before it actually occurs. That is, the likelihood of an anomaly occurring can be predicted by the parameterized programmable circuit, and an action can be initiated, such as control of the signal generating unit to prevent the anomaly from occurring. Therefore, "detection" of an anomaly can also include "prediction" of an anomaly. The trained detection can be part of digital signal processing.

[0059] The results of the digital signal processing, particularly the detection procedure, may be used to control a signal generation unit, such as a DDS core, or additional circuitry in the RF system. The signal generation unit may be used, for example, to adjust the amplitude, frequency, and phase of an output signal that is converted into an RF power signal by a digital-to-analog converter (DAC) and a power amplification stage. As a result, the RF power signal at the output of the RF generator can be provided to the plasma processing system via an impedance matching network. In this way, for example, the results of filtering the input signal from a directional coupler at the output of the RF generator can be directly used to adapt the RF power signal, for example, to stabilize the RF power supplied to the plasma and minimize reflected power from the plasma.

[0060] In particular, the RF range may be 300 kHz to 500 kHz, or 1 MHz to 3 MHz, or 10 MHz to 16 MHz, or 24 MHz to 30 MHz, or 37 MHz to 43 MHz, or 57 MHz to 63 MHz, or 77 MHz to 83 MHz, or 159 MHz to 165 MHz. For example, in the 10 MHz to 16 MHz RF range, the specific frequency of the radio frequency signal may be 13.56 MHz. The aforementioned frequency ranges are particularly used for plasma applications and / or RF generating devices.

[0061] In the processing procedure, at least one input signal is repeatedly processed by the circuit. The circuit may be suitable for executing an algorithm for processing the at least one input signal depending on the configuration of at least one parameter. Thus, the parameter may be a parameter of the algorithm. In the case of filtering, the algorithm may be a filter algorithm, and the parameter may be a filter parameter and / or a filter coefficient. In the case of arc detection, the algorithm may be a pattern recognition algorithm, and the parameter may be a probability of a pattern matching a specific reference pattern. In other words, the probability may represent the similarity between the pattern and the specific reference pattern. The circuit may also be suitable for applying a neural network. In this case, the parameter may be at least one weight of the neural network. The circuit may be programmable in the sense that the configuration of at least one parameter can be changed.

[0062] "Iterative processing" may be understood to mean repeated processing of the same at least one input signal using the same algorithm, but using different configurations. Thus, the configuration of at least one parameter is changed for each processing of the at least one input signal to obtain a respective processing result. The respective processing results may, in other words, be respective results of the algorithm when executed in respective configurations.

[0063] Advantageously, the at least one parameter may be a processing parameter that affects the processing result. If the processing includes filtering of the at least one input signal, the at least one parameter may include at least one filter parameter. For example, the at least one parameter may include an adjustable cutoff frequency and / or a threshold value, etc.

[0064] "Determining at least one parameter result" may be understood as selecting at least one, or exactly one, of the modified configurations that yields the best processing result (and thus is based on the processing result). This allows for optimization of the parameter configurations. "Providing at least one determined parameter result for improved digital signal processing" may also be understood as parameterizing the circuit using the selected configuration.

[0065] The parameter results may include a specific configuration selected from the modified configurations, and the selection may be made depending on the processing results. For example, a criterion, such as a metric or a reference result, may be provided for evaluating each of the processing results. Each of the processing results corresponds to the configuration used to obtain that processing result. As a result, each of the processing results may be used to evaluate, and in particular, rate, the configuration. Thus, the correct configuration corresponding to the best evaluated or best rated processing result may be selected. In other words, the configuration is optimized so that the difference between the corresponding processing result and the reference is minimized. "Optimization of a configuration" may also be referred to as "optimization of at least one parameter," and the at least one optimized parameter (i.e., the selected configuration) may then be used to parameterize (or another) programmable circuit for "in-situ" or "real-time" operation or "application mode" of the radio frequency system. These terms are used interchangeably hereinafter.

[0066] "Evaluation" may be understood as a comparison of the processing result with a standard, such as an ideal processing result. The closer the processing result matches the standard, the higher the evaluation of the corresponding configuration. The evaluation may be performed by a control circuit or by a dedicated part of a programmable circuit. There are various state-of-the-art methods that can be used for the evaluation, such as, for example, statistical methods for signal evaluation, multivariate statistical methods, empirical comparative time series analysis, etc.

[0067] In one embodiment, the at least one parameter result includes at least two different parameter results for different applications of the DSP. Alternatively, or additionally, the DSP and digital signal processing may include filtering that is parameterized by a configuration. The selection of the filtering configuration significantly impacts the reduction of interference in the RF signal. However, one particular configuration may not work for all combinations of RF power levels in a multi-frequency application, nor may it work for all processes performed in the plasma chamber of the plasma system. Therefore, it may be necessary to determine a configuration optimized for different plasma processing conditions. This can be achieved by using multiple parameter results for different applications.

[0068] Similarly, in the case of arc detection, one particular configuration may be optimized only for detecting a particular anomaly, but may not work for other types of anomalies. Thus, detecting several different anomalies may be achieved by sequential and / or parallel processing of the input signal, with a different configuration applied to each process. Thus, the machine learning procedure may be run multiple times for different types of anomalies, so that the features of the anomalies are specific to the different types of anomalies.

[0069] For each application and / or type of anomaly, a different set of input signals can be provided during the training process that can be used separately for the processing procedure, thus resulting in different parameter results. Depending on the sequence of different process conditions in the steps of a plasma processing recipe, for example, when the process conditions are changed during real-time operation, fast switching between different configurations may be advantageous, and therefore different parameter results may be stored in memory for fast switching.

[0070] Within a digital signal processing and / or processing procedure, at least one input signal may be predefined, for example, by providing a previously digitized signal that serves as a training or reference input signal. The previously digitized signal may be a previously recorded signal or an existing signal, for example, a simulated signal. Therefore, an ADC is not necessarily required for processing. However, in real-time operation, the ADC may convert an analog detected RF signal into a digital representation of the signal, hereinafter referred to as a real-time input signal. In training mode, the at least one predefined input signal may be predefined with properties and characteristics desired for the processing procedure and thus "represent" only actual radio frequency signals. However, in real-time mode, the programmable circuit may process a real-time input signal from a detected RF signal instead of the predefined input signal.

[0071] The RF system may include at least one sensor for detecting, particularly measuring, the RF signal used to deliver RF power to the plasma system. The measurement may involve measuring the RF signal, particularly in the form of forward and reverse power and / or RF current and / or RF voltage, at various points within the RF system. It is also intended to adjust the output power of the RF generator, and for this purpose, the RF signal, particularly in the form of an RF waveform at the generator's output, is measured by a sensor. A directional coupler may be used to provide the forward and reverse waveforms to the sensor. Thus, the RF signal may be an analog signal, which is then converted by an ADC with a specific sample rate to provide a digital real-time input signal. Similarly, other sensors, such as phase-amplitude sensors, voltage-current sensors, and DC bias sensors in an impedance matching network, or sensors installed directly in the plasma chamber, such as Langmuir probes or optical sensors, can measure various parameters, such as RF voltage and current, the DC self-bias of the plasma, or the intensity of specific spectral lines in the plasma spectrum. These measurements are used to control adjustable elements to maximize power transfer to the plasma or to characterize the plasma's properties. The measured parameters may be provided as real-time input signals or may be represented as training input signals.

[0072] The programmable circuit may be configured as a programmable logic device, for example, a digital signal processor (DSP), a complex programmable logic device (CPLD), or an FPGA. According to the invention, in particular a programmable circuit may be provided which, at least in part, performs a detection procedure which includes in particular the application of an (artificial) neural network, preferably pattern recognition or pattern matching using a neural network, to identify at least one anomaly characteristic. The detection procedure may provide for the evaluation of measurement data obtained by measuring an input signal, in particular an analog signal.

[0073] In pattern recognition or pattern matching, or preferably in algorithms for identifying at least one abnormal feature in an input signal, and particularly in arc detection, various algorithms can be used: Statistical pattern recognition algorithms, which process the measured data to obtain a set of characteristics such as the mean and standard deviation, correlation coefficient, noise level, deviation amplitude and direction, especially for slowly changing background patterns and fast-changing anomalies. Parameters such as deviation thresholds and correlation coefficients can be used to adapt such algorithms to specific input signals for optimal detection and classification of anomalies. Furthermore, a combination of several features, each with an adjustable weight, may be used to improve the accuracy of anomaly detection and classification. Template matching algorithms, in which the input signal is divided into intervals containing specific structures and then compared to stored templates of anomalies. Parameters for dividing the input signal into meaningful intervals, e.g., periods of steady-state operation and periods with significant deviations from the mean and / or noise level, may be optimized for best detection of features with similarity to the stored templates. Additionally, the scaling factor or scaling factor range used to normalize the input signal so that it can be compared to the templates is also a parameter that can be modified and optimized. Furthermore, variations in the correlation coefficient threshold can be used for further algorithm optimization.

[0074] As an alternative to traditional deterministic algorithms, neural networks can be used to automatically extract features from input signals and classify them. In particular, networks such as the following can be used for analyzing time series and detecting anomalies such as arcs: A feedforward artificial neural network including an input layer, one or more hidden layers, and an output layer, such as a multilayer perceptron (MLP) network with a nonlinear activation function; Convolutional neural networks (CNNs), which include convolutional layers to exploit local correlations in time-series input signals, Recurrent neural networks, especially long short-term memory (LSTM) networks, are well suited to identifying and predicting individual anomalies in input signals that are not noticeable over time.

[0075] When a neural network is used for pattern recognition of an input signal, parameters controlling the neuron weights, the activation function threshold, or the shape of the nonlinear activation function may be parameters that can be varied to optimize the neural network for optimal detection and classification of abnormal features. Backpropagation using gradient methods may be used to train the neural network for optimization of these parameters.

[0076] In training an algorithm or neural network that can be used to process input signals, i.e. in particular in machine learning procedures, a training data set, i.e. a large number of existing input signals, can be used (as at least one input signal according to the invention) and can be generated and provided by using at least one of the following elements: Each of the various anomalies to be detected may need to be represented by a sufficient number of recorded input signals in the training dataset, e.g., 10% or 20%, so that the anomaly training dataset contains both anomaly- and non-anomaly-containing input signals. Certain anomalies may be artificially induced by appropriate parameters in the plasma process or RF power supply system, and may occur more frequently or with higher intensity compared to normal operation. In this way, a large number of input signals with specific anomalies can be recorded in a short period of time. For example, input signals with specific anomalies, evaluated and recorded by an engineer, may be used to automatically generate a set of more similar input signals by scaling, shifting, expanding, compressing, and / or offsetting features, by adding noise with specific characteristics, by adding a slowly varying baseline, or by filtering operations. In this way, batches of input signals with similar anomalies can be generated and included in the training data set. Furthermore, specific anomalies detectable over longer timescales, such as characteristic trends indicative of degradation of specific components, can be generated from the simulation and converted into input signals that are added to the training dataset. Each input signal stored in the training data set can also be completed with its correct classification and / or expected result of processing. In training mode, parameters can be modified to minimize the deviation between the processing result and the expected result for all signals in the training set.

[0077] Particularly in training mode, the at least one input signal may include multiple predefined batches or multiple existing input signals (i.e., previously recorded or, for example, simulated). Each of the input signals may be associated with a radio frequency (RF) signal, i.e., may be specific to or representative of an actual or simulated RF signal. The input signals may also be easily derived from a measured RF signal, for example, by generating a digital representation from the measured RF signal. Furthermore, each input signal may contain a known pattern, and the corresponding reference result may exhibit this known pattern, as described in more detail below. Existing input signals may be thought of as reference or reference signals because they may contain distinctive features and / or provide corresponding reference results. Configuration optimization, i.e., learning the optimal parameterization of the programmable circuit, may be achieved by comparing the reference results with the processed results. The reference results may include a target waveform, a target frequency, or a desired pattern detection result. The comparison may be achieved by using an appropriate metric that indicates the degree of discrepancy between the reference results and the processed results.

[0078] After providing at least one input signal in the form of a batch of input signals, i.e. (at least) a first input signal and a second input signal, in particular at least 10 or 100 or 1000 further input signals, the processing procedure, also referred to as training procedure, may comprise the following processing flow: in a first processing step, processing a first input signal using the first configuration, processing a second input signal using the first configuration, and also processing a further input signal using the first configuration, thereby obtaining a first processing result for the first configuration; a second processing step processing the first input signal using the second configuration, processing the second input signal using the second configuration, and also processing further input signals using the second configuration, thereby obtaining a second processing result for the first configuration; In a further processing step, the first input signal is processed using the further configuration, the second input signal is processed using the further configuration, and also the further input signal is processed using the further configuration, thereby obtaining a corresponding processing result for each of the further configurations.

[0079] Thus, the process is repeated with different parameter configurations, with the configuration being changed for each run. In the training procedure, the change in configuration (and therefore the parameters) can be done by established methods such as gradient descent or backpropagation techniques.

[0080] Within the scope of the present invention, it may be advantageous for the programmable circuit to be configured as a programmable integrated circuit, preferably a digital signal processor (DSP), complex programmable logic device (CPLD), or field programmable gate array (FPGA). Alternatively, the circuit may be configured as a fixed circuit such as an application-specific integrated circuit (ASIC) or another integrated circuit. In general, any computer or controller (e.g., PC, microcontroller, DSP, CLPD, FPGA, etc.) can be used for machine learning, especially machine learning procedures. FPGAs, in particular, are suitable for high-speed, real-time digital signal processing and can accelerate machine learning procedures. Furthermore, the use of programmable circuits, especially FPGAs, enables flexible and automated training. Many applications for radio frequency systems, such as plasma applications, are evolving in various ways, and FPGA solutions have the advantage that future applications requiring adaptive processing, such as as-yet-unknown filter designs, can be handled with the same hardware architecture. In contrast, fixed circuits such as ASICs need to be modified or completely redesigned, which increases the amount of work and costs. The method of the present invention preferably allows for easy addition of new configurations via software upgrades, potentially including more advanced filter configurations, allowing the user to select one of these configurations for a new application.

[0081] Furthermore, within the scope of the present invention, the present invention can further optionally include the following: using at least one determined parameter result to parameterize the same or another programmable circuit; Detects radio frequency signals from radio frequency systems; performing digital signal processing by using at least partially parameterized circuitry to process the detected radio frequency signals; Monitoring radio frequency signals by repeatedly performing detection and digital signal processing, preferably the processing procedure can be referred to as a training procedure used to perform parameterization of the programmable circuit in a training mode of processing using an existing signal as a training input signal and / or can be configured using a parameterization retrieved from memory and possibly previously determined in the training mode, while in particular the monitoring is performed in a real-time mode of processing in which a radio frequency signal is detected and processed as a real-time input signal. This means that the radio frequency system can actually be used in real time and the operation can be switched from training mode to real time mode.

[0082] At a later point in time, the radio frequency system can be switched back from real-time mode to training mode, and the parameterization can be further optimized by iteratively processing additional existing signals, e.g., previously recorded or simulated signals.

[0083] In particular, it is an advantage of the present invention that the programmable circuitry can iteratively process pre-existing training signals to enable both real-time processing of input signals and training of the system by determining the optimal configuration for the real-time processing mode, such as a previously recorded signal or a computer-simulated signal.

[0084] These two features make the RF system highly versatile and allow it to be trained to handle new plasma processing conditions or detect new situations without hardware or algorithm changes. Only the parameterization of the programmable circuit changes. Instead of using an external computer to analyze measured signals and determine an improved configuration, the actual RF system can determine such an optimized configuration by switching it into training mode. In this way, the programmable circuit of the RF system has the characteristics of a machine learning-enabled system or an on-device inference framework, both of which can be adapted to new applications.

[0085] A further advantage of the present invention is that the programmable circuits employed, in particular FPGAs, are much faster, e.g., 10,000 to 100,000 times faster, in performing digital signal processing compared to standard model simulations using standard computers and tools such as Matlab / Simulink.

[0086] The parameter results can include an optimized parameter set, i.e., a respective configuration of at least one parameter optimized using the processing procedure, and the programmable circuit can be parameterized using this parameter set of the parameter results. As a result, the method according to the present invention achieves improved digital signal processing by using highly accelerated optimization of at least one parameter in a first step. In a second step of real-time operation, the processing procedure and the programmable circuit are used, making the selection of a parameter set for digital signal processing faster and smarter. For digital signal processing of the detected radio frequency signal during real-time operation, the programmable circuit can receive at least one real-time input signal converted from the detected radio frequency signal by an analog-to-digital converter.

[0087] Furthermore, the circuitry for optimizing the at least one parameter can be the same as for the final application in real-time operation of a radio frequency system.

[0088] Furthermore, it is conceivable within the scope of the present invention that the at least one input signal processed in the processing procedure is configured as at least one training input signal, and the method further comprises: completing the parameterization of the programmable circuit by configuring at least one parameter using the previously determined parameter results, the configuration of which is preferably subsequently fixed; Switching the processing from a training mode to a real-time mode, in particular in the real-time mode, real-time input signals are processed in real time by the programmable circuit using a fixed configuration, in particular for filtering the real-time signals, preferably the real-time input signals being configured as actual measured signals received from at least one analog-to-digital converter which converts radio frequency signals sensed by at least one sensor into at least one real-time signal; Switching the processing from real-time mode to a training mode, preferably in which the existing, previously fixed configuration is further optimized by running the processing procedure again using at least one additional training input signal as the processed training input signal, and reparameterizing the programmable circuit by configuring at least one parameter using the newly determined parameter result in response to a comparison of the newly determined parameter result with the previously determined parameter result. This allows for further optimization of the existing parameter set. The user can decide whether to switch the system back to training mode to load additional reference input signals and reference results for the existing configuration. The loading or uploading can be performed via an electronic interface or a user interface. Alternatively or additionally, additional reference input signals and reference results can be recorded by the RF system itself and loaded directly into memory in real time for later use in training mode. The old and new sets of parameter configurations are then evaluated, and the better parameter set is subsequently stored in memory. Furthermore, switching from real time mode to training mode allows for training of new situations at a later time. That is, switching to training mode, loading new reference input signals and reference results into memory, and determining and saving an optimized parameter set in memory.

[0089] The user may instruct the RF system via an electronic interface that the training results using the new reference input data are used to optimize an existing parameter configuration, i.e., overwrite the existing parameter configuration with the optimized parameter configuration, or that the optimized parameter configuration is used for a new application and stored in memory as a separate parameter configuration, particularly for the new application.

[0090] According to another embodiment of the method of the present invention, the iterative processing of at least one input signal is performed by iterative processing steps, in each step the same at least one input signal is processed by the same programmable circuit using modified configurations, each processing result being specific to and / or assigned to the modified configuration, in particular the at least one input signal is provided as one or more predetermined reference input signals, for each reference input signal a reference result is provided that represents a desired result of processing the corresponding input signal, and determining at least one parameter result for the optimization of the configuration based on the processing results comprises: The modified configuration is evaluated by comparing each of the processing results obtained for at least one reference input signal with the corresponding reference result, and the evaluation preferably depends on the agreement between the processing result and the reference result, and in particular on identifying deviations between the processing result and the reference result. In other words, the parameters can be optimized by modifying the configuration. Repeated processing of the same at least one input signal may be used, but different configurations in the sense of different parameter values ​​may be applied, resulting in possibly slightly different processing results. Each processing result can be used to evaluate the performance of the specific configuration used to obtain said processing result.

[0091] In particular, the references, reference input signals and / or reference results referred to within the scope of the present invention are also referred to as reference data.

[0092] The evaluation can also be used to provide configuration optimization, in the sense that the parameter results can include a parameter set that results in processing results that minimize deviation from the reference results for all reference input signals. Configuration optimization can also be referred to as parameter optimization. Accordingly, the processing procedure may be referred to as a training procedure and / or an optimization procedure. At least one predetermined reference input signal and the corresponding reference results may be referred to as training data. The determined at least one parameter result can then be verified by performing the processing and evaluation of the processing results again on a second set of reference data (including new input signals and the corresponding reference results). In this way, the problem of overfitting the parameter configuration to the training data is avoided.

[0093] Determining at least one parameter result for optimizing the configuration based on the processing result may be provided within the scope of the present invention and includes: selecting at least one configuration having the highest rating as a result of the at least one determined parameter; In particular, providing at least one determined parameter result for improved digital signal processing, · parameterizing the same or another programmable circuit using at least one selected configuration; wherein preferably, the at least one predetermined reference input signal and corresponding reference result comprise a plurality of predetermined reference input signals and corresponding reference results for different applications of digital signal processing, such that determining the at least one parameter result for optimizing a configuration based on a processing result is performed separately for each application, and preferably, providing the at least one determined parameter result comprises storing a selected configuration in a memory, and parameterizing the same or another programmable circuit comprises selecting a configuration to be used from the stored configurations depending on the application. In other words, different training data sets can be provided for different digital signal processing applications, allowing for switching between parameter sets for different applications. In a further possibility, the at least one input signal is configured as a plurality of existing input signals (also referred to as a "batch of signals"), and preferably the processing procedure comprises iterative processing steps, each processing step comprising: -Process the same input signal one after another, - setting, by the control circuit, the configuration of at least one parameter used in this processing step; The method may further include: After each processing step, the control circuit evaluates the deviation of the processing result from a predetermined reference result; after each evaluation, modifying the configuration of at least one parameter used for a subsequent processing step based on said evaluation; Preferably, the control circuit is electrically connected to the programmable circuit; It is configured separately from the programmable circuit, In particular, the configuration of at least one parameter used is changed for different processing steps to iteratively optimize the improved digital signal processing configuration. The control circuit or control unit may be configured as an additional control processor separate from the programmable circuit, with access to a memory containing training data, e.g., at least one input signal and / or multiple batches of existing input signals and / or corresponding reference results. The existing input signal may be a previously recorded signal, a simulated signal, or the like. The control circuit may also be implemented on the programmable circuit. The programmable circuit may be an FPGA. The provision of the at least one input signal, in particular the selection of training data for the processing procedure (e.g., based on the application to be trained), and / or the modification of the configuration within the processing procedure may not be performed directly by the control circuit, but rather by the programmable circuit itself. For example, for each processing iteration, the control circuit may provide the input signal and / or the modified configuration to the programmable circuit. The control circuit may then perform an evaluation and determine the next change in configuration depending on the evaluation. To accelerate the training procedure, the evaluation may also be implemented at least in part by the programmable circuit.

[0094] Alternatively or additionally, the programmable circuit may be directly connected to the shared memory, or at least a portion of the memory may be implemented by the programmable circuit. In this case, the control circuit may instruct the programmable circuit from which address range of the memory to load the training data. The programmable circuit may then directly access the memory, load and process the training data one after another, and evaluate the processing results. This method has the advantage of significantly speeding up reading from the memory.

[0095] It may also be within the scope of the present invention to configure the processing procedure as a training procedure for iteratively optimizing configurations for different specific situations in a radio frequency system application, so that the parameter results determined for different situations are configured as different optimized parameter sets for these different situations. The different situations may be different digital signal processing applications. The different parameter sets may include different configurations adapted to these different situations. This allows for flexible switching of parameter sets depending on the given situation. Thus, the training procedure enables training of different situations, which can result in parameter sets optimized for different situations.

[0096] In another advantageous embodiment, at least one input signal includes a signal progression over time, preferably instantaneous and historical values, in particular measurements of or derived from a radio frequency signal. A processing result is obtained by processing the signal progression, which indicates the probability of detecting a specific pattern in the signal progression. Thus, not only are the instantaneous values ​​of the input signal considered, but a series of values ​​from a specific time interval prior to the most recent point in time is also processed, allowing the probability of detection, i.e., the expected result, to be calculated. In training mode, the instantaneous and historical values ​​may be predefined and / or previously recorded. However, in real-time mode, the real-time input signal may also include a signal progression over time, preferably instantaneous and historical values, in particular measurements of a radio frequency signal or measurements of a signal derived from a radio frequency signal. In other words, in real-time mode, the processing considers not only the instantaneous measurements of the radio frequency signal, but also a series of values ​​from a specific time interval prior to the most recent measurement point. This allows the processing to be used in different applications, such as probabilistic evaluation.

[0097] At least one input signal, i.e., the training input signal and / or the real-time input signal, may be configured as at least one digital signal. The digital signal may be provided by an analog-to-digital converter that converts the detected analog RF signal into a digital signal. The digital signal may be a digitized representation of the detected RF signal and, therefore, may be treated as a one-dimensional array of data points describing the instantaneous amplitude of the measured RF signal. For measurements that are continuous in time, the array may be further expanded so that the most recent data point represents the most recent measurement. The temporal measurement rate of the measurements, and therefore the temporal measurement rate of the array expansion, may depend on the sampling rate of the digital-to-analog converter. By selecting a time frame of a specific length that always ends with the most recent data point, a series of one-dimensional arrays may be obtained that are updated at the measurement or sampling rate of the digital-to-analog converter by shifting old data points from one end of the array and adding new data points to the other end of the array.

[0098] In particular, real-time measurements may also be needed to identify critical situations and address them before they progress to actual undesirable events. Such methods are particularly useful for handling overvoltage or overcurrent situations in electrical equipment or managing arcing events during plasma processing. As a result, it may be necessary to continuously analyze a stream of measurement data for the presence of specific patterns.

[0099] Advantageously, the method according to the present invention further comprises detecting at least one pattern in the detected radio frequency signals and / or at least one input signal. To improve the reliability of detecting a particular situation, radio frequency signals measured by several sensors can be used in combination, such that at least one pattern is present in each signal at a particular time. For example, successful event detection may require that a particular pattern be detected in two RF signals at the same time or with a certain time delay. Thus, by evaluating at least two signals in combination, false positives due to measurement artifacts or interference, which may occur when evaluating only one signal, can be avoided.

[0100] In particular, "performing digital signal processing at least in part by processing detected radio frequency signals with a parameterized circuit" and / or "performing a processing procedure in which at least one input signal is repeatedly processed with a programmable circuit" respectively include detecting at least one pattern in the detected radio frequency signal and / or at least one input signal. The detected radio frequency signal may be configured as a real-time input signal and thus as a digital signal representation as described above. At least one input signal may also be used in the processing procedure, which may be configured as a training input signal and thus as a digital signal representation as described above. In both cases, the processing may be designed to provide for the detection of at least one pattern, preferably in the form of pattern recognition. Thus, the configuration may be a configuration for pattern detection, particularly a pattern matching or pattern recognition algorithm, and / or the configuration of the weights of individual neurons of a neural network. Pattern matching or pattern recognition algorithms have the advantage of being technically less complex to use than neural networks. However, neural networks can, in some circumstances, detect patterns in noisy or fluctuating signals more reliably than simple pattern matching algorithms. A combination of both is also possible. Thus, the configurable parameters may be configured as parameters of the algorithm or neurons of the neural network. The processing procedure may also be configured as, include, or be part of a machine learning procedure for machine learning and / or training of the neural network, and digital signal processing may be associated with the application of the trained neural network. At least one parameter result for the configuration optimization may further be configured as a training result in the form of a set of neuron weights. The occurrence of an event may be detected and / or predicted based on the detection of at least one pattern.

[0101] It may also be provided that an action is initiated based on the detection or prediction. In particular, the result of pattern detection is a probability of occurrence of the pattern. Additionally or alternatively, the detection and / or detection-based prediction includes determining a probability of occurrence of an event. Initiation of the action may depend on a comparison of this probability with a threshold. This makes it possible in particular to initiate the action before the completion of the event. The action may be a corrective action to counter the occurrence of the event, such as shutting down the RF system.

[0102] In more advanced embodiments, the probabilistic detection or prediction of an event can be based on several conditions. This can be done by monitoring the evolution of the probability of the event occurring over a certain period of time. Rather than simply detecting an event by comparing the instantaneous probability value with a threshold, additional conditions for more reliable detection can include, for example: A probability value that consistently exceeds a certain threshold over a specific period of time. · The average probability value of exceeding a certain threshold over a specific period of time. · Probability values ​​that increase after the probability value exceeds a certain threshold. · Derivative of the probability value within a certain period of time that exceeds a threshold. Correlation of probability values ​​determined for several monitoring signals. Any combination of one or more of the above with further conditions.

[0103] With the help of more complex conditions, the detection of an event can be made more reliable and less dependent on momentary fluctuations in the probability values. Advantageously, by selecting a sufficiently short monitoring period for the probability values, an event can be suppressed before it is completed by initiating an action. Therefore, the training of the system can include parameterization for the earliest possible event detection and / or the most accurate possible event classification. As a result, a higher weighting can be given to the parameterization.

[0104] It may be possible to provide at least one digital signal capable of detecting a particular pattern, particularly in the form of a line array as described above. Preferably, the pattern detection is trained using a training input signal as a digital signal and / or performed in real time using a real-time input signal as a digital signal. The latter allows for identifying an event during operation of the RF system based on the pattern detection. An event may occur in the RF system or plasma processing system and relate to a problem that requires action, such as shutting down the RF system. Furthermore, an event may indicate degradation of a particular component of the RF system or plasma processing system. The event may be reported, for example, by notifying a user via a particular analog or digital signal that may be provided via an electronic or user interface, allowing the user to respond to the situation with additional corrective action or to schedule repairs and / or preventive maintenance for a later time. The specific patterns may include interferences and / or anomalies and / or characteristic changes as described above. These patterns may be associated with specific phenomena in the plasma process, plasma processing system, or RF power supply system. When the RF signal is repeatedly detected by successive measurements, the digital signal represents the continuously detected signal. By continuously processing the digital signal, particularly the resulting sequence of one-dimensional arrays, it may be possible to monitor the temporal changes of the specific patterns.

[0105] In the case of periodic pulse modulation of the RF power signal, abrupt changes in RF power, such as switching between two levels in pulsed mode operation, may be registered as false arcing events by pattern matching algorithms and / or neural networks. In particular, even when evaluating at least two signals in combination, such intentional power changes can be flagged as arcing events. Similarly, an arcing event may also be flagged when the plasma is initially ignited after RF power is turned on.

[0106] To avoid such false positives, the pattern matching algorithm and / or neural network anomaly detection can be turned off for a short period of time ("dead time") whenever the plasma processing recipe or pulse pattern requires a switch in RF power level. In this way, only actual arcing events occurring within a period of constant RF power can be detected.

[0107] Also, by using artificial neural networks, the detection of a specific pattern in its entirety can be learned by optimizing the weights of individual neurons, and using the optimized set of weights, the network can recognize the occurrence of a specific pattern in real time. To detect an event before it is fully completed, an appropriate output function of the neural network can be selected that generates a probability of a specific pattern in a real-time input signal or data stream.

[0108] The same result can also be achieved by calculating the cross-correlation between a reference pattern and the actual measured stream of signal data, where the reference pattern is continuously shifted in time relative to the signal waveform.

[0109] If an event progresses fully, the pattern is recognized with a probability close to 1. However, before that point, the calculated probability is already gradually increasing from 0 or near 0 toward 1. As soon as the probability exceeds a certain user-adjustable threshold or a combination of several of the above conditions, the event can be reported before it actually progresses. In this way, anomalies in the measurement data stream that correlate with undesirable or disastrous events can be flagged before the event occurs. In this case, countermeasures can be initiated to prevent the event from progressing fully and reduce and / or mitigate serious consequences. The probability threshold and further conditions, as described above, can be set individually for each type of event. For example, disastrous events can be treated with a low threshold, while minor anomaly events can be treated with a higher threshold. The advantage of using a low threshold for serious anomalies is that there is more time to implement corrective action before the event progresses fully. On the other hand, a higher threshold allows for more accurate classification of events.

[0110] For example, arcing on the surface of a semiconductor wafer in a plasma processing system can cause serious defects and loss of one or more chips on the wafer. If the arc is strong enough, it can severely disrupt the plasma processing sequence, resulting in the scrapping of the entire wafer. Therefore, for such catastrophic arcing events, a low probability threshold is advantageous so that the event can be suppressed early. An example of the use of this probabilistic approach to suppress an arc before it fully progresses is discussed further below. In another example, a higher probability threshold may be more appropriate for less severe events. In plasma chambers used for film deposition processing, not only the semiconductor wafer but also parts of the plasma chamber are coated with a deposition layer. After processing a large number of wafers, these chamber parts become covered with an increasingly thick layer of deposition material, from which particles may shed. During plasma processing, such particles can cause plasma anomalies that are tolerable to a certain extent. Nevertheless, it is important to accurately classify and measure all particles associated with the event so that preventive maintenance—in this case, chamber cleaning—can be initiated if the number of particles associated with the event exceeds a threshold. In this case, a high probability threshold can separate particle-related events from other anomalies, thus avoiding premature cleaning of the chamber caused by measurements of spurious particle events.

[0111] After detection, classification of the detected anomalies into different anomaly classes may be performed. The classification may be trained in the same way as the detection, i.e., for example, a machine learning procedure may be performed to train the classification, and the detected anomalies may be processed by a programmable circuit to train the classification of the detected anomalies into different anomaly classes. The detection of anomalies in the input signal and classification of these anomalies into different anomaly classes may be related to specific aspects that deviate from normal operation or malfunctions of the RF power supply system, the plasma system, or the plasma process, and may include various signal pre-processing techniques within the detection algorithm or neural network, whereby the signal processing techniques may include, for example, at least one of the following: Subtraction of slowly changing baseline drift and offsets (which may be observed even in normal, non-anomalous input signals) Scaling operation (in time and / or amplitude) before comparing the input signal with the normalized anomaly pattern Shift operations in the time domain, e.g. adding or subtracting a time offset, to adjust for anomalies, especially when evaluating multiple input signals in combination. Calculation of correlation between at least two input signals - Removal of background noise through filtering operations Removal of at least one specific frequency range from the input signal After pre-processing the input signal by one or more of the techniques described above, the actual pattern detection algorithm, including the calculation of probabilities, can be applied to the resulting normalized input signal.

[0112] Implementing an optimized neural network on a programmable logic device, such as an FPGA, is straightforward because all parts of the neural network can be transferred to the corresponding logic components. Similarly, pattern matching algorithms can be easily implemented on an FPGA using shift registers and simple logic operations. Changing the configuration or size of the neural network or modifying the pattern matching algorithm can be done by loading a new configuration into the FPGA without changing the hardware.

[0113] If an application requires the detection of multiple pattern classes, multiple parallel neural networks or pattern matching algorithms can be configured, each optimized for a specific class. It is also possible to use one common neural network or pattern matching algorithm to detect multiple pattern classes, for example, by sequentially evaluating the input signal in separate steps, specifically one for each class. In this case, instead of a single probability value as the network output, an array of probabilities is generated, one for each class. Further pattern class detection can be added later. If necessary, the neural network parameterization or pattern matching algorithm can be optimized by additional training.

[0114] This allows different configurations to be saved for different applications and activated accordingly. In particular, it is possible to quickly switch between different configurations, e.g., for different successive processing steps, one after the other. A teaching mode may then be activated in the installed system to further optimize the existing configuration or learn new situations with additional reference data. Thus, probabilistic pattern recognition in measured signals is also possible by applying parallel (simultaneous) classification to several pattern types, even before the patterns are fully developed.

[0115] Furthermore, an electronic interface may be provided for uploading at least one input signal or an existing configuration to the system of the present invention, particularly to the programmable circuit, or to the control unit, or to the memory. The method of the present invention then uses the uploaded existing configuration to further parameterize the programmable circuit. When an input signal is uploaded to the control unit, the control unit may load the signal into memory before providing it to the programmable circuit. At least one input signal may be configured as multiple input signals, which are stored in the memory after uploading.

[0116] The electronic interface may be configured as a user interface, allowing a user to upload at least one input signal, user input signal, user configuration, control signal for switching the system from training mode to real-time mode, and / or signal using a specific configuration stored in memory for parameterizing the programmable circuit. The uploaded input signal is generally formulated training data and may be supplied to the programmable circuit for digital signal processing directly or via a control unit, or stored in memory directly or via a control unit. The user input signal refers to at least one input signal from a user. The user configuration may be an existing configuration that the user loads into the interface, for example, from cloud or on-site storage, or from a removable data carrier medium that can be connected to the user interface. The user configuration may also be a configuration created by applying the method of the present invention to the user input signal. Generally, the training data may include at least one input signal and a reference result corresponding to each input signal. The training data may be configured as existing (previously recorded or simulated) data, especially for the training procedure, which can be initially loaded into memory and processed one by one (in various iterations) until an optimized parameter set is found. The uploaded configuration can be stored in the same memory and / or a different memory, and / or the uploaded configuration can overwrite a configuration already implemented according to any of the methods of the present invention described above. If the uploaded configuration is stored in the same memory, it can serve as a supplemental / additional parameterization of the programmable circuit. The advantage of storing the uploaded configuration in a separate memory or memory space is that a backup configuration is always available. It also benefits manufacturers by restricting access to configurations other than those created by the user. In this case, loading part of the configuration into a secure and trusted memory space is optional.The advantage of overwriting the configuration of a programmable circuit by uploading another configuration may be a new parameterization of the programmable circuit and a reduction in memory space requirements.

[0117] An electronic interface may be provided for accessing, in particular reading, writing, and / or overwriting, parameter configurations and / or fixed configurations and / or at least one input signal and / or data from the memory and / or the programmable circuit and / or the control unit. However, access via the electronic interface may be restricted, for example, so that reading, overwriting, and / or replacing certain data and / or algorithms or neural networks is prohibited. Thus, some of the data or information stored in the memory and / or the programmable circuit and / or the control unit may be locked, for example, by using encryption methods. In other words, the electronic interface may be used to upload configurations to be stored in the memory for use in a specific application. These configurations may be the result of a training procedure performed on a similar circuit using appropriate training data. In particular, a manufacturer may allow users to upload configurations, but at the same time, protect the basic factory configuration and the algorithm or neural network itself from user access. This has the advantage of allowing a return to the factory configuration if the user is dissatisfied or if the system no longer works with the new configuration.

[0118] Using only the programmable circuit, user interface, memory, and control as a training bench, it is also possible to perform training and determine the optimal parameter set for circuit parameterization. As with the training mode of a complete radio frequency system, already digitized training data is read from memory and provided to the control as input signals for digital signal processing. Parameter changes and evaluation of the processing results are also performed in the same way as for a complete RF system. This training bench system, equipped with a common, commercially available FPGA, can determine optimized parameter results much faster—for example, 10,000 to 100,000 times faster—than standard model simulations using a standard computer and tools such as Matlab / Simulink.

[0119] The optimal parameterization of a typical filtering or pattern matching algorithm used in radio frequency systems can be determined within minutes using the described training bench system, compared to run times of hours or days using Matlab / Simulink simulations.

[0120] For example, even for large neural networks containing thousands of neurons arranged in 500-1000 layers, a training bench system can determine the optimal set of neuron parameters within a few hours. For more complex algorithms or neural networks, and / or very large sets of training input data, determining the optimal configuration of the programmable circuit can be performed on a dedicated high-performance supercomputer.

[0121] Independently of the above possibilities of determining the optimal configuration on the RF system itself, on a training bench system, on a standard computer using simulation tools, or on a high-performance supercomputer, the determined configuration can be uploaded by the control via a user interface into the memory of the RF system and then used to parameterize a programmable circuit capable of processing the RF system's input signals in real time for a specific application.

[0122] Similarly, a configuration determined in one RF system can be exported via the user interface and uploaded to a similar RF system for use in parameterizing the programmable circuits. In this way, the existing configuration of one RF system can be implemented in a second system of the same type, avoiding the need for separate training of the second system. This has the advantage of saving time when setting up a new system, since a copy of an already optimized configuration can be used in multiple systems.

[0123] While unavoidable small variations between systems may require further optimization of the parameterization of a new RF system, this parameter optimization typically converges faster starting from a known good configuration than starting from a default parameter configuration, which is the case for an entirely new application.

[0124] At least one input signal can be configured as a predefined reference input signal, for each of which a desired processing result is provided as a corresponding predetermined reference result. The training data may also include the predefined reference input signals and corresponding predefined reference results.

[0125] To provide processing of at least one input signal, the circuit may execute an algorithm parameterized by at least one configurable parameter of the circuit. An interface, particularly a user interface, may be used to select an appropriate configuration of the parameters before processing begins. Alternatively or additionally, the interface may be used to switch from an application operating mode or a real-time operating mode to a training mode. The processing procedure may be configured as a learning cycle to optimize the configuration. It may also be possible for a user to control the interface to load reference input signals and / or reference results and / or training data into a memory, particularly a memory of the radio frequency generating device, and / or to perform a learning cycle and / or store a new parameter configuration in the memory for later use in the application mode. In the training mode, the memory may be used to simulate a continuous data stream from an analog-to-digital converter from the application mode. Thus, the memory may be connected to a control circuit, which may read at least one input signal from the memory and provide it to the programmable circuit in the training mode. To accelerate processing, the programmable circuit may also read at least one input signal directly from the memory and perform digital signal processing. On the other hand, in the application mode or real-time operating mode, the programmable circuit can be directly connected to an analog-to-digital converter to receive the input signal as a real-time input.

[0126] A further option is to configure the at least one input signal as at least 1,000 or at least 1,000,000 different input signals, each representing at least one waveform associated with a radio frequency signal, and to evaluate the waveforms according to the parameter configuration, whereby each processing result constitutes an evaluation result. This can further improve processing performance. Furthermore, automatic training, for example controlled by the control circuit, can accelerate optimization of the at least one parameter.

[0127] In another advantageous embodiment, the programmable circuit may include logic elements for decomposing an input signal into amplitude and phase using an algorithm, where the algorithm can be parameterized by a modified configuration. Thus, algorithms such as IQ demodulation, the Goertzel algorithm, and the discrete Fourier transform may depend on many parameters. For example, in the case of IQ demodulation, the parameters may be the coefficients of a digital filter. Thus, the configuration of the parameters may be specific values ​​selected for these parameters.

[0128] The evaluation of the modified configuration may be performed by comparing each of the processing results associated with the modified configuration with a predetermined reference result, which may include a predetermined ideal result of the algorithm executed by the programmable circuit. The predetermined ideal result thus serves as a benchmark for the quality of the currently set configuration. Changing the parameter configuration may produce different processing results that can be compared with the reference result. The deviation of the processing result from the reference result indicates the performance of the used configuration, which may be understood as the evaluation of the configuration.

[0129] Yet another aspect of the present invention is an RF measurement system for a system, in particular a radio frequency system, preferably a radio frequency power supply system or a radio frequency generator or a radio frequency plasma system, comprising: at least one sensor for detecting radio frequency signals of the radio frequency system, in particular a directional coupler or a voltage-current (“V / I”) sensor; at least one analog-to-digital converter for converting the at least one detected radio frequency signal into at least one input signal; A circuit, in particular a programmable circuit, for digital signal processing of at least one input signal, which can be provided to be parameterized by at least one parameter result determined according to the method according to the invention. The system according to the present invention thus provides the same advantages as those described in detail with reference to the method according to the present invention. The system may also include an analog anti-aliasing filter and / or a transmission line and / or a matching circuit, in particular an impedance matching circuit. The radio frequency system may be configured, in particular, as a radio frequency power supply system. The radio frequency power supply system may further be configured as a power supply for a plasma system, in particular a plasma power supply system. An analog-to-digital converter may be provided for converting a detected radio frequency signal used in the power supply and detected by a sensor of the radio frequency system.

[0130] The system may further include an electronic interface for uploading an existing configuration to the programmable circuit, to the control unit, or to the memory for parameterization of the circuit. Uploading and / or writing and / or reading using the interface may be performed using data transfer. The electronic interface may be electronically connected to the control unit, the memory, or the programmable circuit to perform uploading using (particularly direct) data transfer. The electronic interface may be configured as a data interface, such as a serial interface, that can be connected to another electronic device, such as a computer. The electronic interface may also execute a computer program to enable data transfer. The computer program may also provide more advanced interface technologies, such as a local area network, a wireless local area network, or Bluetooth. The electronic interface may be accessed via the control unit, the memory, or the programmable circuit.

[0131] The system according to the invention may be characterized by any of the above-mentioned features of the method according to the invention and may further exhibit the advantageous features described.

[0132] Another aspect of the present invention is a data carrier signal carrying at least one parameter result determined according to the method of the present invention. Therefore, the data carrier signal of the present invention provides the same advantages as the corresponding features described in detail for the method and system of the present invention. The data carrier signal can be a signal on a data carrier such as a hard disk or flash memory. Furthermore, the data carrier signal can be implemented as a download, etc. A further possible aspect of the present invention is a computer program including the parameter result determined according to the method of the present invention and / or a computer-readable data carrier storing the computer program and / or the parameter result determined according to the method. The data carrier signal may be provided for performing software upgrades at the user's side, i.e., for parameterizing a programmable circuit using the parameter result determined in the data carrier signal. Alternatively, the data carrier signal may be provided for performing parameterization / configuration reading of the programmable circuit, with the configuration stored in the memory of the RF system or exported, i.e., stored on a removable data carrier medium connected to the user interface of the RF system. The removable data carrier medium can then be connected to a second RF system or training bench system to upload the exported configuration, which can then be used to parameterize a programmable circuit in the second RF system.

[0133] Further advantages, features and details of the invention will become apparent from the following description in which embodiments of the invention are explained in detail with reference to the drawings, in which: In this connection, the features recited in the claims and in the specification may each be essential to the invention individually or in any combination. [Brief explanation of the drawings]

[0134] [Figure 1a]1 shows an embodiment of an RF measurement system 1 of the present invention and a data carrier signal of the present invention. [Figure 1b] 1 illustrates one embodiment of a plasma power system using the RF measurement system 1 of the present invention. [Figure 2] 1 illustrates one embodiment of the method of the present invention for improving DSP in a radio frequency system of the present invention. [Figure 3a] 1 shows an example of the complete evolution of an arcing event without corrective action. [Figure 3b] Shows a similar example to Figure 3a, where arc prevention is initiated when the arc probability exceeds a threshold. [Figure 3c] An example of identifying an actual arcing event during pulsed mode operation is provided. [Figure 4] 1 illustrates one embodiment of the RF measurement system of the present invention in a training mode configuration. [Figure 5] 1 illustrates one embodiment of the RF measurement system of the present invention in application (real-time operation) mode. DETAILED DESCRIPTION OF THE INVENTION

[0135] In the following figures, the same reference numerals are used for the same technical features even in the different embodiment examples.

[0136] FIG. 1 shows an embodiment of the inventive RF system 1, which is a radio frequency generating device, preferably a radio frequency power supply system 1 or a radio frequency generating device or a radio frequency plasma system. The system includes at least one sensor 20, specifically a directional coupler, a voltage / current sensor, or a plasma sensor 32, for detecting the radio frequency signal of the radio frequency system 1. The directional coupler 20 is one type of sensor commonly used to probe radio frequency signals, for example, in power supply circuits. However, other sensors can also be implemented, and FIG. 1a further shows a radio frequency RF power signal transmission line 60 and two signal branches, where the upper branch represents forward power and the lower branch represents reflected power of the RF power signal picked up by the sensor 20. The structure of the inventive system 1 described herein can be suitable for measuring either forward power or reflected power, or both. The structure of the inventive system 1 described herein can be suitable for RF voltage, RF current, reflection coefficient (Γ or “gamma value”), or measured DC self-bias, or any combination of RF current, RF voltage, measured forward or reflected power or both, DC self-bias, or RF phase with reflection coefficient.

[0137] The system 1 further includes at least one analog-to-digital converter (ADC) 30 for converting the at least one detected radio frequency signal into at least one input signal 230. The ADC converts a time-continuous, i.e., continuous-valued, analog signal into a time-discrete, i.e., discrete-valued, digital signal. The analog signal may be a signal picked up by a directional coupler 20, preferably a transmission line 60. The number of samples per second is given by the sampling rate, and the number of distinct values ​​is given by the number of bits of the ADC 30. The system 1 may further include an analog anti-aliasing filter 50 and a matching circuit 40, particularly an impedance matching circuit 40, between the sensor and the ADC 30.

[0138] The system 1 includes a preferably programmable circuit 10 for digital signal processing. The circuit 10 is suitable for digital signal processing of at least one input signal 230. The input signal 230 may be a signal measured during operation of the RF generator, i.e., a real-time signal 230. The circuit is further suitable for digital signal processing of at least one input signal 210 stored in a memory 70. Either the input signal 230 or the stored input signal 210 may be a signal related to a radio frequency signal of a power supply system. Thus, they may be directly measured or simulated signals. The stored input signal 210 may be further processed by a controller or control circuit 80 before entering the circuit 10. The circuit 10 may be further parameterized by at least one parameter result determined according to the method of the present invention described in FIG. 2. The parameterization may occur before the input signal 210 or 230 reaches the circuit 10. The parameterized circuit 10 then outputs a processing result 220 according to the input signal (either from the memory 210 or the real-time mode 230). The processed results 220 may then be passed through the control circuitry 80 and stored in the memory 70 as parameter results 240 .

[0139] FIG. 1a also shows a data carrier signal 240 carrying at least one parameter result 240 determined from the processing result 220. The data carrier signal 240 can be a data package downloadable by an authorized person or any third party. The data carrier signal can then be used to parameterize the programmable circuit 10, particularly the FPGA 10. The processing result 220 used for the parameter result 240 can be determined from at least one of the methods described herein. The system can further include an electronic interface 90 for uploading an existing configuration to the programmable circuit 10, the control unit 80, or the memory 70 for parameterization of the circuit 10. The electronic interface 90 can be electronically connected to the control unit 80, the memory 70, or the programmable circuit 10 for uploading using direct data transfer. Furthermore, the electronic interface 90 can also be used to download a configuration, for example, for transfer to a different RF system for a similar application. The electronic interface 90 can also be a user interface.

[0140] 1b illustrates an embodiment of a plasma power system 5 that uses the inventive RF measurement system 1. In particular, the plasma power system 5 includes an RF generator 2, a matching network 4, and a plasma processing system 3, e.g., a plasma chamber 31 with electrodes to which RF power is supplied.

[0141] The RF generator 2 comprises an RF measurement system 1 including a sensor 20, an analog-to-digital converter 30, a programmable circuit 10, a control unit 80, a memory 70, and an electronic interface 90, also referred to as a user interface 90. Furthermore, the RF generator 2 comprises a signal generation unit 16, e.g., a direct digital synthesis (DDS) circuit and a power amplification stage 23 powered by a DC power supply 21, the signal generation unit 16 being able to be implemented in the programmable circuit 10. It should be noted that within the scope of the present invention, the signal generation unit 16 may also be a separate circuit or chip external to the programmable circuit 10.

[0142] Matching network 4 includes an input sensor 41, e.g., a phase and amplitude detector, an impedance matching circuit 40 that typically includes an inductor and a capacitor that may be variable, a match control unit 42, and optionally an output sensor 43, e.g., a voltage / current sensor or a DC bias sensor. The function of matching network 4 is to match the output impedance (typically 50 ohms) of RF generator 2 to the variable impedance of plasma processing system 3 so that the RF power coupled into the plasma is maximized and the power reflected from plasma processing system 3 is minimized.

[0143] The plasma processing system 3 includes a plasma chamber 31 to which RF power is supplied from a matching network 4. Within the plasma chamber 31, a gas or gas mixture is excited and partially ionized by the supplied RF power. Electrical coupling of the RF power to the plasma can be achieved by connecting the RF power to an electrode, for example, a metal plate opposite a similar plate connected to ground. In this way, the plasma is capacitively coupled to the two electrodes. Alternatively, the RF power can be connected to an inductor positioned around the plasma chamber 31 so that the plasma is inductively coupled.

[0144] Typically, a semiconductor wafer is loaded into the plasma chamber 31 and exposed to an RF plasma for a specified period of time to etch material from or deposit material onto the top layer of the wafer.

[0145] The matching network 4 and the plasma processing system 3 may include sensors 41, 43, 32 that measure RF signals or signals characteristic of properties of the plasma process. Such sensor signals may be connected to dedicated sensor ports in the RF generator 2 and converted to digital representations by an analog-to-digital converter (ADC) 30, e.g., a single-channel and / or multi-channel ADC, and used as input signals for digital signal processing performed by the programmable circuit 10.

[0146] FIG. 2 illustrates an embodiment of a method of the present invention for improving digital signal processing in a radio frequency system, particularly a radio frequency power supply system 1. The method includes providing 101 at least one input signal 210 related to a radio frequency signal of the radio frequency system 1 and executing 102 a processing procedure 310 in which the at least one input signal 210 is repeatedly processed by the programmable circuit 10 using at least one configurable parameter of the processing. The configuration of the at least one parameter can thereby be changed, or specifically modified, for each processing of the at least one input signal 210 to obtain a respective processing result 220. The method further includes determining 103 at least one parameter result for optimization of the configuration (e.g., specifically, an optimized parameter from the changed parameters) based on the processing result 220 and providing 104 the at least one determined parameter result for improved digital signal processing (specifically, for circuit parameterization). In the method of the present invention, the programmable circuit 10 can be configured as a programmable integrated circuit 10, preferably an FPGA.

[0147] The method of the present invention may further include parameterizing 105 the same or another programmable circuit 10 using at least one determined parameter result, the parameterizing 105 including detecting 106 a radio frequency signal of the radio frequency system 1 and performing 107 digital signal processing using at least partial processing of the detected radio frequency signal by the parameterized circuit 10. The method of the present invention may further include monitoring the radio frequency signal by detecting 106 and repeatedly performing 102 digital signal processing. Regardless of the details of the parameterizing 105 of the circuit 10, the processing procedure 310 of the programmable circuit 10 may be configured as a training procedure for performing the parameterizing 105 of the programmable circuit 10 in a training mode of processing using the input signal 210 as an existing (e.g., previously recorded or simulated) training input signal 210, while the monitoring is performed in a real-time mode of processing, where a radio frequency signal is detected and processed as a real-time input 230. The at least one input signal 210 processed in the processing procedure 310 can be configured as at least one training input signal 210. The method of the present invention can further include finalizing the parameterization 105 of the programmable circuit 10 by configuring at least one parameter using a previously determined parameter result, with the configuration then being fixed. The method of the present invention can further include switching processing from a training mode to a real-time mode, in which the real-time input signal 230 is processed in real time by the programmable circuit 10 using the fixed configuration, particularly for pattern detection of the real-time input signal 230. Preferably, the real-time input signal is configured as a measured signal received from at least one analog-to-digital converter 30, which converts a radio frequency signal sensed by the at least one sensor 20 into at least one real-time input signal. The method of the present invention can further include switching processing from the real-time mode to a training mode. In the training mode, the existing, previously fixed configuration can be further optimized by re-executing the processing procedure 310 using at least one additional training input signal 210 as the processed training input signal 210 to newly determine at least one parameter result. In response to a comparison of the newly determined parameter results with the previously determined parameter results, a reparameterization of the programmable circuit 10 is performed by configuring at least one parameter using the newly determined parameter results.

[0148] The repeated processing of the at least one input signal 210 can be performed through repeated processing steps, in which the same at least one input signal 210 is processed by the same programmable circuit 10 using a modified configuration of at least one parameter, so that each processing result 220 is unique to and / or assigned to the modified configuration. The at least one input signal 210 can be configured as one or more predetermined reference input signals 257, for each of which a reference result representing a desired outcome of processing the corresponding input signal 210 is provided. The determination 103 of the at least one parameter result for the configuration optimization based on the processing results 220 can further include evaluating the modified configuration by comparing each of the processing results 220 obtained for the at least one reference input signal 257 with the corresponding reference result. The evaluation depends on a match between the processing result 220 and the reference result or a deviation between the processing result 220 and the reference result. Determining 103 at least one parameter result for optimizing a configuration based on processing results 220 can further include selecting at least one configuration with the highest evaluation as the at least one determined parameter result. Criteria results can be provided by a user, such as by uploading them using user interface 90. Thus, a user can define at any time what criteria will be used for the evaluation. Input signals, and thus training data, can also be defined by a user and uploaded, for example, via user interface 90 for that reason. Providing 104 at least one determined parameter result for improved digital signal processing can further include parameterizing 105 the same or another programmable circuit 10 using the at least one selected configuration.Preferably, the at least one predetermined reference input signal 257 and corresponding reference result comprises a plurality of predetermined reference input signals 257 and corresponding reference results for different applications of digital signal processing, such that determining the at least one parameter result for optimizing a configuration based on the processing result 220 is performed separately for each application. Providing 104 the at least one determined parameter result may further comprise storing the selected configuration in a memory, and parameterizing the same or another programmable circuit 10 comprises selecting a configuration to be used from the stored configurations depending on the application. At least one input signal 210 can be configured as a plurality of pre-existing input signals 210. The pre-existing signals can be previously recorded or, for example, simulated. The processing procedure 310 can include repeated processing steps, each processing step including processing the same input signal 210 one after the other and setting, by the control circuit 80, the configuration of at least one parameter used in that processing step. The method of the present invention can further include, after each processing step, evaluating, by the control circuit 80, the deviation of the processing result 220 from a predetermined reference result and, after each evaluation, modifying, based on said evaluation, the configuration of one parameter used for the next processing step. Thus, the control circuit 80 can be electrically connected to the programmable circuit 10 and configured separately from the programmable circuit.

[0149] The configuration of the at least one parameter used can be changed for different processing steps in order to iteratively optimize the configuration for improved digital signal processing.

[0150] The processing procedure 310 can be configured as a training procedure to iteratively optimize the configuration for different specific situations in the application of the radio frequency system 1, so that the parameter results determined for different situations are configured as different optimized parameter sets for these different situations.

[0151] The at least one input signal 210 can preferably include instantaneous and historical values, in particular signal progression over time from measurements of radio frequency signals. The processing result 220 is obtained by processing the signal path and can indicate the probability of detecting a particular pattern in the signal path.

[0152] An electronic interface 90 may further be provided for uploading at least one input signal 210 to the programmable circuit 10, whereby the at least one input signal 210 may be configured as multiple input signals 210, which may be stored in memory after uploading.

[0153] The at least one input signal 210 can be configured as at least 1,000 or at least 1,000,000 different input signals 210, each representing at least one waveform associated with a radio frequency signal, where the processing can be aimed at evaluating the waveforms according to the parameter configuration, and each processing result 220 can include a result of said evaluation.

[0154] An existing configuration can also be uploaded via electronic interface 90 to programmable circuit 10, control unit 80, or memory 70, among others. Subsequently, parameterization of programmable circuit 10 uses the existing configuration. The uploaded existing configuration can be stored in the same memory 80 and / or a different memory. The uploaded existing configuration can overwrite any of the configurations calculated using the method of the present invention, or can be stored in memory in addition to other configurations calculated using the method of the present invention or previously uploaded and stored in memory.

[0155] Evaluation of the modified configurations may be performed by comparing each of the processing results 220 associated with the modified configurations with predetermined reference results, which may include predetermined ideal results of the algorithms executed by the programmable circuit 10.

[0156] FIG. 3a shows an example of an arc event observed in two different sensor signals, a first sensor signal 402 and a second sensor signal 403. The bottom curve represents RF power 404. The x-axis represents time 405. In this case, the arc need only be monitored; no preventative measures are initiated, and the arc may proceed fully until the discharge is complete. After training a pattern matching algorithm or neural network for this particular case of arcing, the algorithm may provide a probability value p(t) 406 (shown in the upper curve of FIG. 3a, representing probability 401) that increases from zero at time tstart to one at time tend. The first signs of an arc event are evident at time tstart, and the event is completed at time tend. After the arc event, the probability value p(t) 406 may rapidly decrease to zero again. The first sensor signal 402 may be configured as a current sensor signal, and / or the second sensor signal 403 may be configured as a voltage sensor signal.

[0157] A similar situation is shown in Figure 3b. As in Figure 3a, the top curve represents probability 401, the two curves below represent first sensor signal 402 and second sensor signal 403, and the bottom curve represents RF power 404. The x-axis represents time 405. In contrast to Figure 3a, the probability value p(t) 406 is a function of the threshold p threshold 407, and the probability value p(t) 406 is compared to this threshold p threshold 3a, the probability value p(t) 406 of an arc event p(t) 406 may increase gradually starting at time tstart. When the probability value p(t) 406 exceeds the threshold p thresholdUpon reaching 407, a temporary reduction in RF power 404 may be triggered. This action prevents the arcing event and allows the plasma process to return to normal operation. Notably, the spike in the first sensor signal 402 in FIG. 3a is significantly reduced, which may mean that the arcing has been suppressed.

[0158] FIG. 3c illustrates an example of a method for efficiently suppressing false positive detections. The x-axis represents time 405. In this example, the RF power output 404 may be modulated with a pulse pattern that regularly switches between two RF power levels. By disabling arc detection in the time intervals t1a...t2a, t1b...t2b, t3a...t4a, t3b...t4b, t5a...t6a, t5b...t6b, t7a...t8a, and t7b...t8b, spikes in the transitions between low and high RF power levels are ignored, allowing only two arc events unrelated to RF power switching to be detected. In this example, the two arc events can be classified into two different classes: arc event class A 408 and arc event class B 409 (as is evident from their different shapes).

[0159] In the case of periodic pulse modulation of the RF power signal 404, sudden changes in the RF power 404, such as when switching between two levels in pulsed mode operation, may be registered as false arcing events by pattern matching algorithms or neural networks. In particular, even when evaluating at least two signals in combination, these intentional power changes may be flagged as arcing events.

[0160] To avoid such false detections, the detection of anomalies via pattern matching algorithms or neural networks can be turned off for a small period ("dead time") whenever the pulse pattern requires a switch in RF power level. In this way, only actual arcing events occurring within a period of constant RF power 404 can be detected. An example of the use of such dead time is shown in Figure 3c.

[0161] 4 illustrates an embodiment of the RF measurement system 1 of the present invention configured to operate in a training mode, such as a training bench system. The operating mode 253 of the system 1 can be set to training mode via the user interface 90.

[0162] In the following, the term "signal processing circuitry" referred to by reference numeral 17 refers to a portion of programmable circuit 10 that is specialized to perform signal processing, in particular digital signal processing. Programmable circuit 10 may include further circuitry, for example evaluation circuitry 18, that is used in training mode to determine the deviation between processed result 220 and a reference result. The further circuitry used during real-time mode may be implemented within programmable circuit 10, but may be deactivated in training mode.

[0163] In training mode, the power amplifier stage 23, the impedance matching network 4, and the connections of the system 1 to the plasma processing system 3 are not required. Instead of real-time input signals provided by the RF generator 2, the matching network 4, or sensors in the plasma processing system 3, the system 1 may process existing reference input signals 257 stored in the memory 70 or uploaded via the memory 70. Additionally, reference results 256 may be stored in the memory 70 or provided via the user interface 90. These reference results indicate expected results of signal processing of the reference input signals 257. Additionally, parameter configurations 254 that can be used to parameterize the signal processing circuit 17 may be stored in the memory 70 or provided via the user interface 90 as user parameter sets 251. Digital signal processing algorithms or, for example, neural network circuit layouts 255 that can be implemented in programmable circuit 10 may also be stored in memory 70. Typically, uploading algorithms and circuit layouts 255 via the user interface and implementing them in programmable circuit 10 is restricted to manufacturers and is not accessible to general users of system 1.

[0164] In training mode, controller 80 can be configured to read user parameter sets 251, user criteria results 252a, and user criteria input signals 252b from user interface 90 and store them in memory 70. Additionally, controller 80 can be configured to read stored parameter sets 254 from memory 70 and parameterize programmable circuit 10 accordingly.

[0165] Additionally, the controller 80 can be configured to read the reference input signals 257 from the memory 70 or the user interface 90 and provide them to the signal processing circuit 17 as input signals 210. Alternatively, the signal processing circuit 17 can directly access the memory 70 and directly read the parameter configuration 254 and training data, i.e., the reference input signals 257 and the reference results 256, thereby accelerating the reading and processing of the training data.

[0166] 4 can be implemented within the programmable circuit 10 or as a software algorithm within the controller unit 80. The result of the evaluation, i.e., the optimized parameter set 254, can be stored in the memory 70 by obtaining it from the programmable circuit 10 and writing it to the memory 70 via the controller 80 or by direct memory access from the programmable circuit 10.

[0167] 4, it is noted that at least part of controller unit 80 and / or at least part of memory 70 can be implemented in programmable circuit 10. This has the advantage that the reading of reference input signal 257 and reference result 256 from memory 70, and the subsequent signal processing of reference input signal 257 and evaluation of processing result 220 against reference result 256, can be significantly accelerated.

[0168] 5 illustrates an embodiment of the RF measurement system 1 of the present invention in application (real-time operation) mode. In application mode, the signal processing circuit 17 is parameterized using an existing parameter set 254 stored in memory 70 from a previous training session or uploaded to memory 70 via user interface 90. Depending on a process recipe 258, which may be provided via user interface 90, a dedicated parameter configuration 254 may be loaded into the signal processing circuit 17 for each recipe step from memory 70 or from the user interface 90 via controller 80. The signal processing circuit 17, thus parameterized, processes input signals in real time that are continuously provided from an RF system, such as the RF generator 2, the impedance matching network 4, or sensors in the plasma processing system 3, and digitized by the ADC 30.

[0169] Depending on the algorithm or neural network implemented in the programmable circuit 10, the results of the signal processing may be used to control a signal generation unit 16, e.g., a DDS core, or further circuitry 19. The signal generation unit 16 is used, for example, to adjust the amplitude, frequency, and phase of a signal that is converted into an RF power signal by a digital-to-analog converter (DAC) 22 and a power amplification stage 23. The resulting RF power signal at the output of the RF generator 2 may then be provided to the plasma processing system 3 via an impedance matching network 4. In this way, for example, the results of filtering the input signal 210 from the directional coupler 20 at the output of the RF generator can be directly used to adapt the RF power signal, for example, so that the RF power delivered to the plasma is stabilized and the reflected power from the plasma is minimized.

[0170] Additionally, signal processing circuitry 17 may be configured to monitor input signal 210 for patterns indicative of specific events or undesirable conditions, such as a high-voltage discharge or secondary plasma within plasma processing system 3, or degradation of hardware components related to plasma processing, such as particle buildup within the plasma system, or a drift in plasma impedance that indicates the need for cleaning or other preventative maintenance of plasma chamber 31. Such events can be flagged 259 to a user via user interface 90 or acted upon directly by further circuitry 19 within programmable circuit 10, which can trigger corrective action 250. For example, in the event of an arcing event or high-voltage discharge, or if signal processing determines that such an event is likely to occur, programmable circuit 10 may initiate a reduction or complete shutdown of RF power for a period of time. This can be achieved, for example, by reducing the amplitude of the RF signal generated in signal generation unit 16, by disabling DAC 22, or by interrupting the connection between DAC 22 and power amplifier stage 23 for a period of time.

[0171] The circuit layout of programmable circuit 10 or the algorithms used by programmable circuit 10 may also be stored in memory 70, and different circuit layouts 255 may be implemented in programmable circuit 10 depending on the application.

[0172] As mentioned above for the training mode, even in the application mode, at least a portion of the controller 80 and / or a portion of the memory 70 can be implemented within the programmable circuit 10. This has the advantage that switching between different configurations of the signal processing circuit 17 can be significantly faster.

[0173] The above-described embodiments are merely examples of the present invention, and it is understood that the features of the embodiments can be freely combined within a technically reasonable range without departing from the spirit of the present invention. [Explanation of symbols]

[0174] 1 System 2 RF generator 3. Plasma Processing System 4. Matching network 5 Plasma Power System 10 circuits 16 Signal Generation Unit 17 Signal processing circuit 18 Evaluation circuit 19 More Circuits 20 sensors 21 DC power supply 22 Digital-to-Analog Converter (DAC) 23 Power Amplifier (PA) 30 Analog-to-Digital Converter (ADC) 31 Plasma Chamber 32 Plasma Sensor 40 Impedance matching circuit 41 Input Sensor 42 Match Control Unit 43 Output sensor, V / I sensor 50 Anti-aliasing Filter 60 Transmission Line 70 memory 80 Control circuit 90 Electronic Interfaces, User Interfaces 101 Providing at least one input signal, first method step 102 Performing the second method step 103 Decision, third method step 104 Providing at least one determined parameter result, fourth method step 105 Parameterization, fifth method step 106 DETECTION, 6TH METHOD STEP 107 Performing the seventh method step 210 (stored / existing) input signal 220 Processing Results 230 (real-time) input signals 240 Parameter Results 250 Corrective Actions 251 User Parameter Set 252a User Criteria Results 252b User Reference Input Signal 253 Operating Mode 254 parameter sets 255 Circuit Layout 256 Standard results 257 Recorded measurement data sets, reference input signals 258 Process Recipes 259 Event Flag 310 Processing Procedures, Machine Learning Procedures 401 Probability 402 first sensor signal 403 Second Sensor Signal 404 RF power 405 Time 406 p(t) 407p threshold 408 Arc Event Class A 409 Arc Event Class B

Claims

1. 1. A method for machine learning detection of at least one anomaly in a plasma system, particularly an RF powered plasma processing system, the method comprising: providing at least one input signal (210) having at least one anomaly characteristic related to an analog signal of a power supply system (1) of the plasma system and / or another characteristic of the power supply system (1) and / or the plasma system and indicative of the anomaly in the plasma system; 11. The method of claim 10, further comprising: executing a machine learning procedure (310) in which the at least one input signal (210) having the at least one anomaly feature is processed by a programmable circuit (10) to train the detection of the anomaly in the plasma system.

2. 2. The method of claim 1, wherein the programmable circuit (10) is configured as a programmable integrated circuit (10), preferably a digital signal processor (DSP), a complex programmable logic device (CPLD), or a field programmable array (FPGA).

3. The programmable circuit (10) is characterized in that it at least partially executes a detection procedure that includes in particular the application of a neural network, preferably pattern recognition or pattern matching using said neural network, or an algorithm to identify said at least one anomalous feature, said machine learning procedure (310) comprising: performing said processing of said at least one input signal (210) using said detection procedure and at least one configurable parameter of said detection procedure, in particular the weights of said neural network or the parameters of said algorithm, said processing of said at least one input signal (210) being changed, in particular modified, for each processing of said input signal (210) to obtain a respective processing result (220); The method of claim 1 , further comprising determining at least one parameter result as a training result of the machine learning based on the processing results (220), the parameter result including, in particular, the selection of the modified configuration.

4. 4. The method according to claim 3, characterized in that the machine learning procedure comprises iterative processing steps, in which the same at least one input signal (210) is processed by the programmable circuit (10), but using modified configurations having different parameters, to obtain the respective processing results (220) assigned to the used configuration, and an evaluation of the modified configurations is performed by comparing each of the obtained processing results (220) with a reference result.

5. 5. The method of claim 4, characterized in that the evaluation of the modified configurations depends on matching the processing results (220) with the reference results, and the at least one configuration having the highest evaluation is selected from the modified configurations as the determined at least one parameter result.

6. The method of claim 4 , wherein the baseline result is a predetermined indication of the at least one abnormality.

7. 7. The method according to claim 1, wherein each of the at least one input signal (210) is related to a radio frequency signal used to supply power to a plasma system and / or to another characteristic of the power supply system (1) and / or the plasma system, and the anomaly is specific to an arc occurring in the plasma system or specific to the occurrence probability of the arc.

8. 7. The method of claim 1, wherein the machine-learned detection of the at least one anomaly is used in arc detection and / or arc prevention and / or arc management.

9. the machine learning procedure (310) provides for an iterative determination of the configuration of at least one parameter of a detection procedure for said detection, in particular of at least one weight of a neural network or of at least one parameter of an algorithm, in particular of said neural network or said algorithm implemented at least in part by said programmable circuit (10), said determined configuration being subsequently used for said detection procedure in the field operation of said power supply system (1); The method according to claim 1 , further comprising the step of outputting warning information when the abnormality is detected by the detection procedure.

10. Providing the at least one input signal (210) comprises:

7. The method according to claim 1, further comprising recording and converting analog signals of the power supply system (1) in the form of radio frequency signals and / or other characteristics of the power supply system and / or the plasma system, and providing digital representations of the analog signals and / or the other characteristics to the programmable circuit (10) as the at least one input signal (210).

11. 7. The method according to claim 1, wherein a detection procedure for the detection of the at least one anomaly is implemented in the programmable circuit (10) by combining functional blocks of the programmable circuit (10) for performing calculations in real time.

12. The method described in claim 11, characterized in that the implemented detection procedure is repeatedly used in real-time calculations during the machine learning procedure (310) for the processing of the at least one input signal (210).

13. 7. The method according to claim 1, wherein the machine learning (310) is configured as a training procedure of a parameterization of the programmable circuit (10), such that one and the same programmable circuit (10) or another circuit (10) having the same parameterization can be used for the detection of the at least one anomaly during field operation of the power supply system (1), in particular for arc detection and / or arc prevention and / or arc management.

14. In the on-site operation, when the at least one abnormality is detected, the following operation is performed: Shutdown of at least a part or all of the power supply system (1); Shutdown of the power supply system (1) without restarting the power supply system (1); A temporary shutdown of the power supply system (1), followed by a restart of the power supply system (1); A temporary reduction in the output power of the power supply system (1); Temporarily modifying the output frequency of the power supply system (1), A temporary modification of at least one modulatable element in the impedance matching network (4); 14. The method according to claim 13, characterized in that at least one of the following is initiated individually or in combination:

15. 7. The method according to claim 1, wherein the detection of the anomaly comprises a probabilistic detection, in particular for detecting an arc before it occurs in a plasma processing system, preferably for determining the probability of the occurrence of the arc and / or for arc prevention.

16. 1. A system for detecting at least one anomaly in a plasma system, particularly an RF powered plasma processing system, comprising: at least one sensor (230), in particular a directional coupler and / or a voltage and current sensor, for detecting at least one radio frequency signal of a power supply system (1) of the plasma system and / or another characteristic of the power supply system (1) and / or of the plasma system; at least one converter (30), in particular an analog-to-digital converter (30), connected to said sensor (230) for converting said at least one detected radio frequency signal and / or said further characteristic into at least one input signal (210); a programmable circuit (10) for digital signal processing of said at least one input signal (210), A system, characterized in that the programmable circuit (10) is trained for the detection of the at least one anomaly according to the method of any one of claims 1 to 6.

17. A data carrier signal (240) carrying said at least one parameter result (240) determined according to the method of any one of claims 1 to 6.

Images