Adaptive Filter for Lithography Metrology Noise Suppression

The patent implements a multi-stage adaptive filtering system where filter coefficients are dynamically adjusted based on real-time measurement conditions. The system transitions between different filter stages (first, second, and third stages) with increasing aggressiveness, selecting the appropriate stage based on the detected noise level and rate of change. This dynamic adaptation allows the system to maintain low tracking error during stable conditions while effectively suppressing noise during turbulent conditions, resolving the contradiction between noise suppression and tracking accuracy.

The system changes filter parameters (coefficients and stages) based on the measured noise characteristics. By monitoring the rate of change and magnitude of measurements, the system adjusts the filter aggressiveness parameter, switching between conservative (first stage), moderate (second stage), and aggressive (third stage) filtering. This parameter adaptation enables the system to optimize the trade-off between noise suppression and phase lag compensation under different operating conditions.

The system dynamically switches between conservative and aggressive filtering modes based on real-time conditions. During periods of low noise and stable measurements, the system uses the first stage filter with conservative coefficients to minimize phase lag and maintain tracking accuracy. When noise levels increase or measurement variability exceeds thresholds, the system transitions to second and third stages with more aggressive coefficients, thereby achieving effective noise suppression without permanently sacrificing tracking accuracy.

The filtering approach applies periodic reassessment of noise conditions and adaptive adjustment of filter stages. The system continuously monitors measurement characteristics and periodically transitions between filter stages, applying the appropriate level of filtering aggressiveness. This periodic adaptation ensures that the system maintains optimal performance by switching from conservative to aggressive filtering as conditions warrant, and vice versa.

The filtering system is segmented into three distinct stages, each with specific coefficient ranges and application conditions. The first stage uses conservative coefficients for normal conditions, the second stage uses moderate coefficients for intermediate conditions, and the third stage uses aggressive coefficients for high-noise conditions. This segmentation allows the system to achieve high adaptability by selecting the appropriate stage, while keeping each individual stage relatively simple and well-defined.

Different filter coefficients and stages are applied locally based on the specific noise conditions detected. Rather than using a single complex adaptive algorithm, the system applies simple, well-defined coefficient sets locally appropriate to each operating condition. The controller selects which local filter configuration to apply based on real-time measurements, achieving high adaptability through conditional selection rather than through complex continuous adaptation.