Frequency Selective Iterative Learning Control for Stage Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Precision assembly systems, such as semiconductor wafer exposure systems, face challenges in achieving precise stage movement due to positioning errors caused by component rigidity and mechanical vibrations, which affect the quality and throughput of semiconductor wafer manufacturing.
Innovation Solution
An iterative learning control (ILC) system is employed to reduce positioning errors by generating an error signal, transforming it into the frequency domain, selecting specific frequency components, and using these components to generate a filtered error frequency representation, which is then used to create a current control signal to correct stage movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional feedback control loops and feed-forward compensation schemes are used to control stage movement, then basic positioning control is achieved, but positioning precision remains insufficient for nanometer-level requirements
Solution Approach 1:
The patent employs iterative learning control that uses feedback from previous iterations to improve positioning accuracy. The control system measures positioning errors from prior cycles and uses this information to generate corrected control signals for subsequent iterations, progressively reducing errors to achieve nanometer-level precision while maintaining system robustness through continuous learning and adaptation
Solution Approach 2:
The control system performs preliminary error measurement and analysis during idle periods or between production cycles. By pre-processing error data and preparing corrected control signals before the next positioning operation, the system can apply optimizations in advance, ensuring high precision without compromising real-time control robustness
2Manufacturing precision
If the control system repeatedly performs the same stage movements for imaging multiple identical ICs on the same wafer, then repetitive positioning errors occur due to mechanical vibrations and component flexibility, but conventional control methods cannot eliminate these recurring errors
Solution Approach 1:
The iterative learning control system is self-improving, automatically measuring its own positioning errors and generating corrected control signals without external intervention. The system uses its measurement capabilities to identify repetitive errors from mechanical vibrations and flexibility, then self-corrects by applying learned compensation patterns in subsequent iterations, achieving high repetitive positioning accuracy while maintaining relatively simple hardware architecture
Solution Approach 2:
The control system dynamically adjusts control parameters based on measured positioning errors from previous iterations. By changing control signals in response to observed errors rather than using fixed parameters, the system can compensate for repetitive disturbances caused by mechanical vibrations and component flexibility, achieving improved positioning accuracy without adding complex mechanical structures
3Measurement precision
If feedback control loops are added to reduce positioning errors, then control accuracy improves, but system complexity and computational load increase
Solution Approach 1:
The iterative learning control system creates a virtual model of positioning errors by measuring and storing error patterns from previous iterations. Instead of implementing complex real-time control algorithms, the system uses this error copy to generate corrected control signals for subsequent operations, achieving high measurement precision with relatively simple control structure by leveraging historical error data
Data Source
AI summary
Methods and control systems are provided for controlling stage position errors based, in some embodiments, on a selection of frequency components in a stage position error signal. An error frequency representation of a position error signal may be generated in the frequency domain and filtered by selecting one or more desired frequency components. The filtered error frequency representation can then be manipulated according to a control law and transformed back into the time domain to generate a current control signal. The current control signal can then be used to adjust the position of the stage to reduce positioning error.