Model for fault detection in a plasma system- Higher order harmonic impedance analysis

• Main Authors: Szu-Po Wang, 王思博
• Other Authors: Bing-Hung Chen
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/34562221936176515738

碩士 === 國立東華大學 === 電機工程學系 === 93 === There is a bottleneck in the future semiconductor manufacturing due to the physic sensitivity limitation of process sensors used in the conventional monitoring system which monitor and control single-variate. Although multi-variate statistics approach can enhance sensitivity to establish control chart, it is tedious to collect “golden samples” to setup model. Therefore an effective, simple, elegant methodology for monitoring micro-change in plasma is needed. This research is based on the principle of relation between micro-change of plasma electrical property and sheath variation. This variation can induce incident electromagnetic waveform deformity. We design high sensitivity voltage and current probes to measure the deformed waveforms, then derive first five order voltage, current and phase by fast Fourier transformation to get high order harmonic impedances represented plasma condition. This kind monitor system is non-intrude and has no impact to plasma itself. By using Taguchi experimental design to establish fault samples, and Baye’s theorem to classify the source of fault, we has successful detected faults of RF power with 5% fluctuation and working pressure with 10% fluctuation. Equipments we made in this experiment are relatively simple and economic to collect high order harmonic signal. The manifest feature is the trace back mechanism for the fault reason provided by this methodology.