Abstract

Optical Emission Spectroscopy (OES) is a non-intrusive plasma diagnostic technique that can be used to measure the chemical changes in a plasma that is in- creasingly being considered for monitoring and control of plasma etch processes. In the practice of collecting OES data for plasma etching, it is inevitable that noise is included in the measurements. The existence of noise can destroy signals or at least make the identi¯cation and interpretation of signal patterns unreliable, hence appropriate ¯lter- ing and pre-processing of the data is needed prior to application of automated feature extraction and analysis techniques. In the absence of a priori knowledge of the noise characteristics and system bandwidth the selection of an appropriate noise suppression ¯lter bandwidth is a challenging problem. This paper explores the characteristics of the noise inherent in OES measurements and proposes a systematic method for establish- ing a suitable noise ¯lter bandwidth based on the auto-correlation and cross-correlation analyses of the ¯ltered signals and their residuals.