即時預測模型於晶圓測試回收良率之應用

Abstract

晶圓測試完成後，為了減低因誤判造成低良率晶片重新架機之重工機率，於是在晶片退出測試設備前會針對每一個不良晶粒執行立即重測，期望能救回因為測試問題所造成的誤判，來提昇晶圓測試良率並省去事後重工的架機時間。由於不良的原因可能來自於晶圓廠製程異常所產生的產品問題、測試設備不穩定，或是因測試作業不當所造成；如此一來，不管是產品問題或測試誤判，在此模式下都會進行晶圓重測。若重測後提昇之良率不敷測試成本，便間接造成測試產能的浪費。然而大部分測試工程師所關心的，除了降低測試問題造成良率異常外，更想直接透過立即重測來救回誤判之晶粒，彌補測試問題並提昇晶圓良率，期望在重測後的良率比未重測前能夠提高。因此為了提升晶圓良率，往往只注意良率的變化，而未考量重測實際所帶來的重測效益。 本研究收集案例公司於99年之晶圓測試資料、測試成本與產品利潤之相關資料以供研究分析。論文首先探討案例中影響重測回收良率之因素，並透過迴歸分析法預測執行重測可回收之良率。並依此預測結果來比較測試成本及重測利益的關係，決定是否值得在晶圓退出測試設備前就執行立即重測，希望能為案例公司帶來最佳效益，為此論文之主要貢獻。 實驗的分析顯示，本論文所提出的預測方法能為案例公司增加測試利潤。此模型在晶圓測試業有其應用的價值，可以增加生產的效益，提高產業的競爭力。

On-line re-probing function of failed dies is used to increase the 1st yield in order to reduce the probability of wafer re-loading for off-line re-testing. It will perform re-testing before wafers are unloaded from the prober when the result of the 1st yield is low. This is for the purpose of saving the setup time in off-line re-testing. Most of the product engineers are interested in reducing off-line re-testing and increasing the wafer yield by using the on-line re-probing function of failed dies. However, it wastes testing productivity if the re-testing cost is higher than re-testing benefit. In this study, we collected wafer testing and cost related data from one of Taiwan’s testing facilities in 2010. We analyzed the factors of wafer recovery yield and established the regression model to predict the recovery yield for the on-line re-probing function of failed dies. As a result, the yield prediction model is used to evaluate the re-testing cost and benefit. It’s recommended to perform on-line re-probing of failed dies if the re-testing benefit is greater than the cost. The recovery yield prediction model will help testing facilities to establish a cost-effective re-testing mechanism in order to reduce cost and increase productivity of testing equipments. This is the main contribution of this study. According to the test results of empirical studies, the prediction model can be applied to increase net profit. Thus, it is valuable for the testing industry to enhance production efficiency and improve the company’s competitiveness.