Vacuum Valve Feed-Forward Pressure Control for Process Chambers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vacuum valve systems in semiconductor fabrication face challenges in achieving rapid and reliable pressure regulation due to delayed feedback and unpredictable gas fluctuations, leading to reduced fabrication quality.
Innovation Solution
A two-stage learning method for a controller to derive and apply feed-forward controlling parameters, including calibration and process calibration steps, to optimize vacuum valve operation for precise and swift pressure control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If delayed feedback control is used for pressure regulation, then the control system is simple to implement, but the pressure regulation speed and reliability deteriorate
Solution Approach 1:
The patent applies preliminary action by implementing a feed-forward control mechanism that uses a trained neural network model to predict and compensate for pressure variations before they occur. The controller is trained offline using historical process data to learn the relationship between process parameters and pressure changes, enabling it to proactively adjust valve positions in response to predicted pressure fluctuations rather than reacting to delayed feedback signals.
2Adaptability or versatility
If traditional feedback control is used, then the control logic is simple, but the adaptation to gas fluctuations and process conditions is poor
Solution Approach 1:
The patent applies parameter changes by utilizing a neural network model that dynamically adjusts control parameters based on learned relationships from training data. The controller transforms static control logic into an adaptive system that modifies its behavior based on process conditions, gas fluctuations, and chamber characteristics, achieving high adaptability through data-driven parameter optimization rather than fixed control algorithms.
3Manufacturing precision
If rapid pressure regulation is implemented, then fabrication quality improves, but the control system requires advanced learning algorithms
Solution Approach 1:
The patent resolves this contradiction by performing the complex learning and adaptation work in advance during an offline training phase. The neural network model is trained using historical process data to establish optimal control strategies for various process conditions. During actual fabrication operations, the pre-trained model executes rapid predictions and control adjustments without requiring real-time complex computations, thus achieving both high fabrication quality and operational simplicity.
Data Source
Figure 1
Figure 2
AI summary
Method for providing controlling parameters for controlling a vacuum valve (10) to control the pressure in a process chamber (1) by means of the vacuum valve (10), wherein the method comprises performing a valve calibration step by providing a calibration chamber having an upstream gas inlet and the vacuum valve to control a downstream gas outlet flow, providing defined gas flow into the calibration chamber, running a valve calibrating cycle with defined variation of a valve position of a valve closure of the vacuum valve, measuring a chamber pressure in the calibration chamber, modelling a controlling behavior for the vacuum valve based on the measured chamber pressure and the related variation of the valve position and deriving respective calibration controlling parameters for the vacuum valve based on the controlling behavior. The method comprises performing a process calibration step by providing a process chamber comprising at least an upstream gas inlet and the vacuum valve to control a downstream gas outlet flow, measuring an actual pressure in the process chamber, performing a pre-process step with the process chamber (1) with given process target pressure and given process timing, controlling the actual pressure in the process chamber by means of the vacuum valve to provide the process target pressure with the process timing and deriving process controlling parameters based on the controlling. Further, the process controlling parameters and/or the calibration controlling parameters are provided to control a process cycle.