郭瑞祥2006-07-252018-06-292006-07-252018-06-292001http://ntur.lib.ntu.edu.tw//handle/246246/3033晶圓允收測試資料係指晶圓在完成所有製 造程序後，測試結構所量測的電性測試參數。 此類參數的分析能夠快速的評估整體製程狀 態，並提供了製程異常狀態的警訊以及產品元 件的電性特性。然而由於晶圓允收測試資料是 在生產線末端量測而得，他們具有多機台及時 序錯亂的雙重效應，因此通常不容易直接偵測 出其異常趨勢。為了提昇異常趨勢的偵測速 度，我們利用到生產線上某部機台的製程有速 移時，晶圓允收測試資料的平均值及變異數都 會增加之特性，同時監看晶圓允收資料的平均 值、變異數及製程能力指標，因此發展了一套 結合Shehwart 管制圖，指數加權移動平均圖 (Exponentially Weighted Moving Average)與指 數加權移動Cpk 圖的晶圓允收監看系 統:SHEWMAC桹灨敮經由模擬分析與晶圓廠資料實 證分析，驗證本研究之SHEWMAC 法較傳統的 Shewhart-EWMA 法更能快速偵測製程異常速 移 。Under the effects of multiple-stream and sequence-disorder, process change caused by one machine at an in-line step may result in changes in both the mean and variance of end-of-line wafer acceptance test (WAT) data sequence. To speed up trend detection of WAT data without resorting to an intensive computing power, an end-of-line SHEWMAC scheme is proposed, which combines a Shewhart, an exponentially weighted moving average (EWMA), and an exponentially weighted moving Cpk (EWMC) charts for jointly monitoring the mean and variance of wafer lot average sequence from WAT data. In view of the wide ranges of process conditions and low volume of each product in a foundry fab, a data normalization technique is adopted to aggregate data of similar products and a new design method is developed to generate a robust set of scheme parameters. Simulation and field data validation show that SHEWMAC is superior to the combined Shewhart-EWMA scheme in shift detection speed and is complementary to the in-line SPC.application/pdf228386 bytesapplication/pdfzh-TW國立臺灣大學工商管理學系晶圓允收測試資料統計製程管制指數加權移動估測法Wafer Acceptance TestStatistical Process ControlExponentially Weighted Moving Averagereporthttp://ntur.lib.ntu.edu.tw/bitstream/246246/3033/1/892213E002125.pdf