經濟部智慧財產局員工消費合作社印製 414997 A7 _____B7_ 五、發明說明（） 發明領域： 本發明係有關於一種半導體製程中的品質控制方法， 特別是有關於一種利用統計計算方法為基礎，在半導體製 程令執行品質管制及系統化的管理方法。 發明背景： 在半導體製程令’為了提高半導體晶圃的製造品質、 降低製造成本、縮短生產週期等等目的，資訊管理系統整 合自動化的製造是現今半導體廠所要研究的重要課題。一 個完整的電腦整合自動化系統大致包含了例如生產排程系 統、品質管理系統、監控系統、和自動化介面系統等等。 由於半導體產品的不斷推陳出新，各個製造流程和細節也 更趨複雜，因此自動化系統中就必需有一個良好的生產排 程系統和管理系統；而為了元件的製造品質’更必需有一 套品質管理系統和監控系統加以控管。 在半導體元件持績往輕、薄、短、小和高密度的方向 發展時’其元件的品質管制顯得更加重要。因為元件更為 精密時，其靈敏度也將大為提高，自然其缺陷的容忍度也 跟者降低了’而為了要充分監測及反應半導體元件的缺 陷，在半導想…有許多監測系統和儀器加以控制β例 如，在晶片製造過程中’會在許多較特殊的站别設立檢查 本紙張尺度適用中圉國家標準（CNS)A4規格（210 X 297公餐） — — — — — — —I-1 *t— — I — 11 I 訂· 1----I--I (請先閱讀背面之注意事項再填寫本頁) 414997 經濟部智慧財產局員工消費合作社印製 Α7 Β7 五、發明說明（） 點’透過此些檢查站即可確定晶圆上是否有額外的缺陷， 或是製造過程上是否出現異常，以便及早發現及處理。例 如晶®上之光阻顯影後之外觀檢查（After Develop Inspection，ADI)即為其中之_個檢査項目。 在傳統的電腦整合製造系統中雖然有品質管理系統來 測試元件的可靠度和品質分析等步驟，但是其測試方式較 為固定。也就是說’系統將考量整體之製造品質來進行檢 測及分析的步驟，例如上面所述的製程中設立檢查站，只 要有待測物經過此檢查站，都必需進行檢查測試的動作， 以確保製程的礪利進行。 然而’隨者製程的複雜化，其檢查站的設置也隨之增 加。但是有許多檢查測試站的設置，雖然可發現製程的異 常，但對元件的可信賴度影響並不大，並不需要針對每個 待測件進行檢測，如果經過每一個檢査站的待測件都必需 執行檢查的話，不但生產的週期會增加，產能受到影響， 而且製造成本也會因而增加•對於此種情形，將需要—套 品質管理系統’可以容易的解決此項問題，以便減少製造 成本，增加產能。 發明目的及概述： 3 本紙張尺度適用中國國家標準（CNS>A4規格（210x 297公釐） ----------- I J -----—--訂------I--線 1 • - <請先閲讀背面之注意事項再填寫本頁) 經濟部智慧財產局員工消費合作社印製 414997 A7 _______B7______ 五、發明說明（） 赛於上述之發明背景中，傳統的電腦整合製造系統在 進行品質管制時，如果製程中具有許多測試步驟，將無法 自動的根據產品之品質狀況調整檢測的頻率，因而產生了 增加的產品製造週期和製造成本，所以本發明之主要目 的’即針對此問題點’以產品之穩定性為統計取樣的基礎， 作為製程尹檢測步驟是否進行的依據，以進行產品之品質 管制。 本發明的另一目的在以系統管理的方法，動態的調整 檢測的取樣頻率’以符合實際狀況。 根據以上所述之目的，本發明提供了一種以統計取樣 為基礎來進行製程中品質管制的系統。本系統將依附在原 本的生產管制系統（Manufacturing Executive System，MES) 中，此生產管制系統將控制所有在製品以進行最正確的生 產指引及追蹤；同時也具備了品質控制系統，以便執行製 造批號的檢測程序，並進而分析元件之信賴度 (reliability)。之後，生產管制系統將產品的檢測結果送到 製程統計分析器（SPC analyzer)中，此分析器中具有資料 庫’可儲存生產管制系統傳來的產品檢測結果，並可根據 此檢測結果以統計方法分析取樣的頻率’再將得到的取樣 頻率存入取樣頻率資料庫（sampling rate database)中。然 4 本紙張尺度適用中國國家標準（CNS)A4規格（210 X 297公3 ) <請先閱讀背面之江意事項再填寫本頁) 裝—----丨丨訂丨11 —--線 414997 A7 B7 經濟部智慧財產局員工消費合作社印製 五、發明說明（） 後，整個作業核心的背景伺服器電腦系統（backgr〇und server)將會讀取取樣頻率資料庫中的資料’再依據此取樣 資料而動態的調整控制生產管制系統中的產品檢測頻率。 如果製程步驟或者相關製程參數有所更動時，操作者 亦可經由圖形化的使用者介面將取樣頻率輸入到取樣頻率 資料庫中，以進行取樣測試頻率的更冑，達到動態變更檢 測頻率的目的。 經由此系統所管制的製程步驟中，如果待測件經過檢 測站時’系統將判斷此產品的取樣頻率，進而決定此製造 批號是否需要檢測。而如果待測件的測試結果顯示品質不 穩定時，也會反應到本系統中而自動的調高測試頻率，因 此本系統可以有效的利用檢驗測試設備，達到系統化的管 理及降低成本和維持產品品質的目的。 圓式簡單說明： 本發明的較佳實施例將於往後之說明文字中辅以下列 圓形做更詳細的闡述，其中： 第一圈為依照本發明之統計取樣以進行品質管制的系 統結構方塊圖； --------震------- - 訂--------線 (請先閱讀背面之注意事項再填寫本頁) 414997 A7 經濟部智慧財產局員工消費合作社印製 B7 五、發明說明（) 第二圊為依照本發明之統計取樣結果以判斷檢驗步驟 是否執行的流程示意圖；及 第三圖為依照本發明之統計取樣結果動態調整取樣頻 率之示意囷 發明詳細說明： 由於半導體製程的複雜性與多樣化，如果不適時的運 用系統化和結構化的資訊管理流程，則將難以即時掌握生 產的狀況。在傳統之生產系統中，雖然從產品開發、生產 及製程的規劃、產品測試及檢驗等等流程，已逐漸經由電 腦化管理控制所取代，但製造廠多元化的需求仍時常造成 自動控制上的盲點’因而無形中消耗了製造成本。本發明 即是根據產品在製程中進行測試及檢驗時，生產系統無法 動態調整測試頻率的缺點’而發展出以產品的穩定性作為 統計取樣的基準，進而判斷產品測試的頻率，以便適度的 控制產品品質的系統化管理方法，以下即以圖示說明本發 明之詳細内容》 參閱第一圖，此結構方塊圖說明了依照本發明之系統 進行製程中的統計取樣，並進行品質管制的流程。方塊1〇〇 本紙張尺度適用令國國家標準（CNS)A4規格（210 X 297公釐） (請先閱讀背面之注意事項再填寫本頁) 裝!!丨訂-------線 經濟部智慧財產局員工消費合作社印製 414997 A7 _____ B7 五、發明說明（） 稱之為生產管制系統（Manufacturing Executive System， MES)，此系統可提供模組化的分散式管理，例如生產排程 系統（scheduling syStem)(未顯示於圖中）可提供正確的生產 指引給生產機台和操作人員，如那些貨（I〇t)要到那一站， 生產條件為何等相關資訊。而製造程序（pr〇cess pr〇cedUre)110則描述產品的製造流程，不同產品其製造流 程亦不同。製造批號追縱（l〇ttrack;ing)120則可進行泰本資 料之收集處理、在製品（WIP)追蹤、或操作者的指引等。 於生產管制系統1 〇 〇中，尚且包含了品質管理系統 (Quality Management System)，此系統將收集在製品的製程 及生產結果的董測資料，和檢驗測試的資料，以便進行產 品之品質分析等工作。另外，在生產管制系統1 00中亦包 含有監控系統進行製程設備之監控，及自動化介面系統進 行設備控制和資料傳輸等工作。 在此生產管制系統100中，方塊130代表製造之批號 將依一般之製程步驟進行處理，並且跳過檢測之步驟，此 類批號亦稱之為未取樣之製造批號；而方塊140則代表製 造之批號將經過檢測之步驟，也就是某一個製程步驟執行 完畢後，將執行檢測步驟，再接著下一個製程步驟。此類 批號稱之為取樣之製造批號。至於如何區分製造批號是否 i紙張&度剌中@ S家標準（CNS)A4規格⑵ο X 297公釐） ---------t l· . ----1 I--訂·!---I--線 (請先閱讀背面之注意事項再填寫本頁) 414997 經 濟 邨 智 慧 財 產 局 員 工 消 費 合 作 社 印 製 A7 B7 五、發明說明（） 要經過檢測步驟，則將由背景伺服器400加以控制。 一個完整的自動化系統不僅僅由一群不同的電腦藉由 網路連結而已’其作業核心亦需要功能完整的分散式系統 加以支援，背景伺服器400即為執行此類分散式系統軟體 之主電腦。同時，其中一個故障的電腦只會影響局部的運 作，並不會導致全線停擺。 當生產管制系統1 00檢測完取樣之製造批號時，會將 檢測之資料和結果送到製程統計分析器（SPC anaIyzer)200 中，製程統計分析器200中具有資料庫，可储存每筆測試 的資料’並依-般的統計方法計算檢測值，進而分析測試 取樣的頻率。 之後，製程統計分析器200根據計算的結果而選取適 當的取樣測試頻率’並將此數㈣存到取樣頻率資料庫· 中。背景词服器綱即可藉由取樣領率資料庫300取得產 品之取樣頻率，並將其回饋到生產營制系統⑽中。 取樣頻率資料庫300中的資姐介 的資枓亦可經由使用者介面 5〇0加心㈣輪人，錢^介面州即為—般之圈形 化使用者介面（GUI),當製程條件改 雙或有其它的特殊狀況 本紙張尺度剌+關家鮮（CNS>A4胁（210 (諳先閱讀背面之汶意事項再填寫本頁} 裝·-------訂---------線 414997 A7 __ ___B7 五、發明說明（） 時’操作者即可根據實際狀況經由此介面調整取樣頻率， 而不必受制於原始的製程統計分析器200。同時操作者也 可藉由此使用者介面500而得知不同產品的取樣頻率變 化’進而加以監督’或甚至更正製程統計分析器2〇〇的分 析規則。 第_圖中所示之系統架構的操作情形則為：當新的產 品導入生產時，製程統計分析器200將設定最初的取樣頻 率為1(取樣頻率1即為100%的檢驗），並存入取樣頻率資 料庫300中。背景伺服器400取得此數值之後，即可輸入 到生產管制系統1 〇 〇中，然後此製造批號之產品將全部經 歷檢驗步驟。然而，經過數次的測試之後，產品品質漸趨 穩定，由製程統計分析器200可經分析得知此狀況，於是 依照分析的結果自動將取樣頻率降低並储存於取樣頻率資 料庠300中，此時背景伺服器400將動態的依據取樣頻率 資料庫300中的數值而改變產品的測試頻率。 相反的，當一個穩定的產品於測試時顯示其結果有轉 | 壞的傾向時，製程統計分析器2〇〇亦將分析出此狀況而自 ^ 動將取樣頻率提高，於是此製造批號將受到嚴密的監控， | 直到品質狀況再度回到穩定狀態為止》 產 局 消 費 合 作 社 印 製 本紙張尺度適用中因囡家標準（CNS)A4規格（210 X 297公芨） '-1 --------I i l· . --------訂--I--I ---線 *- (諳先閱讀背面之注意事項再填寫本頁) 414997 A7 B7 五、發明說明（） 參閱第二圖’此圖中所顯示的為生產管制系統1〇〇中 製造程序110之製程安排流程圖。在複雜的半導體製程步 驟中，通常都會在其中設置檢驗站，以確保製程之品質。 於圖中的方塊220、230、260和270分別為製程步驟〗' 2、 3、及4，當晶0經由製程步驟^2(方塊220、230)的執行 之後’將進入方塊240來判斷是否要進行檢測的步琢，而 此判斷的標準即為上述之背景伺服器400所取得此產品之 取樣頻率。經過判斷之後，如果條件不符合，則直接跳過 檢測步驟250而執行製程步驟3、4，否則將進八檢測步称 25 0執行檢驗，然後才執行製程步驟3、4。 現參閱第三圖，此圖中所顯示的係依照本發明之製程 穩定度而調整產品檢驗頻率的示意圊。上面曾描述本發明 之系統可自動的依照製程穩定度而調整檢測取樣頻率，當 新的產品導入生產或製程穩定度極低時，取樣頻率將提高 (最高為1，即100%檢驗），而當製程穩定度昇高之後，製 程統計分析器200即可由統計計算而得知，進而將較低的 取樣頻率存入取樣頻率資料庫300中供背景伺服器400取 用。 於此圖中的最左邊一列為製程穩定度，最右邊一列則 為檢測取樣頻率，而中間所示的方塊係為製程中之待測 本紙張尺度適用辛國國家標準（CNS)A4規格（210x297公釐） (請先閱讀背面之注意事項再填寫本頁) ^一^------訂--I I---1 1^. 經濟部智慧財產局員工消费合作社印製 A7 414997 __ B7__ 五、發明說明（） 件。當於此檢測站之前的製程穩定度很低或新導入之產品 時（圖中之第一橫列）’為了發現製程之問題以維持製造品 質，取樣頻率將設定為1，此時所有的產品經過此檢測站 均將接受檢驗（以斜線面積表示）；當製程穩定度漸趨穩度 時（圖中第二橫列之中等狀態），取樣頻率即降為1 /2，即為 連續兩個製造批號（lot)中只選擇其一進行檢驗；而製程穩 定度變得更高之後（圖中之第三橫列），取樣頻率更降到 1/3，即為連續三個製造批號中平均只選擇其一進行檢驗； 但是如果製程穩定度又降到中等狀態時（圖中之第四橫 列），則取樣頻率將又回昇至1 /2的狀態》 經過本發明上述之詳細說明後，製程中品質控制的取 樣檢測頻率即可藉由產品的成熟度和穩定性而經由統計計 算方法加以控制。同時’此系統不僅可以系統化的方式自 動管理檢測步驟之執行’以節省成本，更可經由動態的輸 入而根據實際之製程狀況調整取樣頻率。 如熟悉此技術之人員所瞭解的，以上所述僅為本發明 之較佳實施例而已’並非用以限定本發明之申請專利範 圍；凡其它未脫離本發明所揭示之精神下所完成之等效改 變或修飾，均應包含在下述之申請專利範圍内。 各紙張尺度適用中國國家標準（CNS)A4規格（210 X 297公爱） · lit --------t---------0 I (請先閱讀背面之注意事項再填寫本頁) 經濟部智慧財產局員工消費合作社印製Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs and Consumer Cooperatives 414997 A7 _____B7_ V. Description of the Invention () Field of the Invention: The present invention relates to a method of quality control in semiconductor manufacturing processes, and particularly to a method based on the use of statistical calculation methods in semiconductors. The process order implements quality control and systematic management methods. Background of the Invention: In order to improve the manufacturing quality of semiconductor wafers, reduce manufacturing costs, shorten production cycles, etc., in semiconductor process orders, integrated and automated manufacturing of information management systems is an important subject to be studied by semiconductor factories today. A complete computer integrated automation system roughly includes, for example, a production scheduling system, a quality management system, a monitoring system, and an automated interface system. Due to the continuous innovation of semiconductor products, and the various manufacturing processes and details are becoming more complicated, a good production scheduling system and management system must be included in the automation system; and for the manufacturing quality of components, a quality management system and The monitoring system controls it. As semiconductor components continue to develop toward light, thin, short, small, and high density, the quality control of their components becomes even more important. Because when the component is more precise, its sensitivity will be greatly improved, and its tolerance for defects will naturally be reduced accordingly. In order to fully monitor and respond to the defects of semiconductor components, there are many monitoring systems and instruments in the semiconductor. Control β For example, during the wafer manufacturing process, 'there will be set up in many more special stations to check the paper standard applicable to the national standard (CNS) A4 specification (210 X 297 meals) — — — — — — —I- 1 * t— — I — 11 I Order · 1 ---- I--I (Please read the notes on the back before filling this page) 414997 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Α7 Β7 V. Description of the invention () Click 'through these inspection stations to determine if there are additional defects on the wafer or if there are abnormalities in the manufacturing process for early detection and processing. For example, after Develop Inspection (ADI) on the photoresist on the wafer is one of the inspection items. In the traditional computer integrated manufacturing system, although there is a quality management system to test the reliability and quality analysis of components, the test method is relatively fixed. In other words, the system will consider the overall manufacturing quality to perform the inspection and analysis steps. For example, an inspection station is set up in the process described above. As long as the object to be tested passes this inspection station, inspection and testing operations must be performed to ensure that The sharpening of the process proceeds. However, the complexity of the 'follow-up' process has increased the number of checkpoints. However, there are many inspection and test station settings. Although abnormalities in the process can be found, the reliability of the component is not significant. It is not necessary to perform inspection for each DUT. If the DUT passes through each inspection station, If inspection is required, not only the production cycle will increase, the production capacity will be affected, but also the manufacturing cost will increase. For this case, a quality management system will be needed to easily solve this problem in order to reduce manufacturing costs. To increase production capacity. Purpose and summary of the invention: 3 This paper size applies to Chinese national standard (CNS > A4 size (210x 297 mm)) ----------- IJ -------- Order ----- -I--line 1 •-< Please read the notes on the back before filling out this page) Printed by the Employees' Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 414997 A7 _______B7______ V. Description of Invention () In the above background of the invention, traditional During the quality control of the computer integrated manufacturing system, if there are many test steps in the manufacturing process, it will not be able to automatically adjust the detection frequency according to the quality status of the product, thus resulting in increased product manufacturing cycles and manufacturing costs. The purpose 'that is to address this problem point' is to take the stability of the product as the basis for statistical sampling, and use it as the basis for the inspection of the manufacturing process to perform product quality control. Another object of the present invention is to dynamically adjust the detection sampling frequency 'in accordance with the actual situation by using a system management method. According to the above-mentioned object, the present invention provides a system for performing quality control in a process based on statistical sampling. This system will be attached to the original Manufacturing Executive System (MES). This production control system will control all work in progress for the most accurate production guidance and tracking; it also has a quality control system to perform manufacturing lot numbers Testing procedures, and further analysis of component reliability (reliability). After that, the production control system sends the test results of the products to the process statistical analyzer (SPC analyzer). This analyzer has a database 'can store the product test results from the production control system, and can use the test results to collect statistics. The method analyzes the sampling frequency ', and then stores the obtained sampling frequency in a sampling rate database. Of course, 4 paper sizes are applicable to Chinese National Standard (CNS) A4 specifications (210 X 297 male 3) < Please read the Italian matter on the back before filling out this page) Loading ------ 丨 丨 Order 丨 11 ----- Line 414997 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. 5. After the invention description (), the background server computer system (backgrund server) of the entire operation core will read the data in the sampling frequency database. Based on this sampling data, the product inspection frequency in the production control system is dynamically adjusted and controlled. If the process steps or related process parameters are changed, the operator can also input the sampling frequency into the sampling frequency database through the graphical user interface to update the sampling test frequency and achieve the purpose of dynamically changing the detection frequency. . In the process steps controlled by this system, if the DUT passes the inspection station, the system will judge the sampling frequency of the product, and then determine whether the manufacturing lot number needs to be inspected. And if the test result of the DUT shows that the quality is unstable, it will also be reflected in this system and the test frequency will be automatically increased. Therefore, this system can effectively use inspection and testing equipment to achieve systematic management and reduce costs and maintenance. The purpose of product quality. Brief description of the circle type: The preferred embodiment of the present invention will be described in more detail in the following explanatory texts with the following circles, where: The first circle is a system structure of statistical sampling for quality control according to the present invention Block diagram; -------- Earthquake --------Order -------- line (Please read the precautions on the back before filling this page) 414997 A7 Intellectual Property Bureau, Ministry of Economic Affairs Printed by employee consumer cooperative B7 V. Description of the invention (2) The second diagram is a schematic flow chart of the statistical sampling results according to the present invention to determine whether the inspection step is performed; and the third diagram is a dynamic adjustment of the sampling frequency according to the statistical sampling results of the present invention Schematic and detailed description of the invention: Due to the complexity and diversification of the semiconductor process, if a systematic and structured information management process is not applied at an appropriate time, it will be difficult to grasp the status of production in real time. In traditional production systems, although the processes of product development, production and process planning, product testing and inspection, etc. have been gradually replaced by computerized management and control, the diversified needs of manufacturing plants still often cause automatic control. 'Blind spots' thus virtually consume manufacturing costs. The present invention is based on the disadvantage that the production system cannot dynamically adjust the test frequency when the product is tested and inspected in the manufacturing process, and develops the stability of the product as a reference for statistical sampling, and then judges the frequency of product testing for appropriate control. The method of systematic management of product quality, the following is to illustrate the details of the present invention. Refer to the first figure. This structural block diagram illustrates the process of statistical sampling in the process and quality control according to the system of the present invention. Box 100 This paper size is applicable to the national standard (CNS) A4 specification (210 X 297 mm) (Please read the precautions on the back before filling this page) Assembling !! Ordering ----- Line Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 414997 A7 _____ B7 V. Description of the Invention () It is called the Manufacturing Executive System (MES). This system can provide modular decentralized management, such as production scheduling The system (scheduling syStem) (not shown in the figure) can provide the correct production guidelines to the production machine and operators, such as which goods (Iot) will go to the station, what are the production conditions and other related information. The manufacturing process (prOcess prOcedUre) 110 describes the manufacturing process of the product, and the manufacturing process is different for different products. Manufacturing batch number tracking (ing track; 120) can be used to collect and process Thai materials, work in process (WIP) tracking, or operator's guidance. In the production control system 1000, a quality management system is also included. This system will collect data from the process of production process and production test results, as well as inspection and test data, for product quality analysis, etc. jobs. In addition, the production control system 100 also includes a monitoring system for monitoring process equipment, and an automated interface system for equipment control and data transmission. In this production control system 100, block 130 represents the manufacturing lot number will be processed in accordance with the general manufacturing process steps, and the detection step is skipped. Such lot numbers are also called unsampled manufacturing lot numbers; and block 140 represents the manufacturing lot number. The batch number will go through the test step, that is, after the execution of a certain process step, the test step will be performed, and then the next process step. Such lot numbers are called manufacturing lot numbers for sampling. As for how to distinguish whether the manufacturing batch number is paper & degree @ S 家 standard (CNS) A4 specification ⑵ X 297 mm) --------- tl ·. ---- 1 I--Order · ! --- I--line (please read the precautions on the back before filling out this page) 414997 Printed by the Consumers' Cooperative of the Economic Village Intellectual Property Bureau A7 B7 V. Description of the invention () After going through the detection steps, the background server will be used 400 to control. A complete automation system is not only supported by a group of different computers through a network connection. The core of its operation also needs to be supported by a fully functional decentralized system. The background server 400 is the host computer that executes such decentralized system software. At the same time, one of the malfunctioning computers will only affect local operations and will not cause a shutdown of the entire line. When the production control system 100 detects the sampled manufacturing lot number, it will send the test data and results to the process statistical analyzer (SPC anaIyzer) 200. The process statistical analyzer 200 has a database that can store the data of each test. Data 'and calculate the detection value according to the general statistical method, and then analyze the frequency of test sampling. After that, the process statistical analyzer 200 selects an appropriate sampling test frequency 'according to the calculated result and stores this number in the sampling frequency database. The background server can obtain the sampling frequency of the product through the sampling lead database 300 and feed it back to the production management system ⑽. The resources of the sister-in-law in the sampling frequency database 300 can also be added through the user interface 5000. The interface state is the same as the circled user interface (GUI). When the process conditions Change the double or have other special conditions This paper size 剌 + Guan Jiaxian (CNS > A4 threat (210 (谙 Please read the Wenyi matters on the back before filling out this page}) Install · ------- Order --- ------ line 414997 A7 __ ___B7 5. When the invention is explained (), the operator can adjust the sampling frequency through this interface according to the actual situation, without being restricted by the original process statistical analyzer 200. At the same time, the operator can also Through this user interface 500, it is known that the sampling frequency of different products changes 'and then be monitored' or even corrects the analysis rules of the process statistical analyzer 200. The operation situation of the system architecture shown in Figure _ is: When a new product is introduced into production, the process statistical analyzer 200 will set the initial sampling frequency to 1 (sampling frequency 1 is 100% inspection) and store it in the sampling frequency database 300. The background server 400 obtains this value It can then be entered into the production control system 1 〇, then the products of this manufacturing batch will all go through the inspection step. However, after several tests, the product quality gradually becomes stable. The process statistical analyzer 200 can learn this situation through analysis, and then automatically follow the analysis results. The sampling frequency is reduced and stored in the sampling frequency data 庠 300. At this time, the background server 400 will dynamically change the testing frequency of the product according to the value in the sampling frequency database 300. In contrast, when a stable product is being tested, When the results show a tendency to turn | bad, the process statistical analyzer 2000 will also analyze this situation and automatically increase the sampling frequency, so this manufacturing lot number will be closely monitored, | until the quality status returns again Up to a stable state "Printed by the Bureau of Industry and Consumer Cooperatives, the paper size is applicable to the family standard (CNS) A4 specification (210 X 297 meters) '-1 -------- I il ·. ---- ---- Order--I--I --- line *-(谙 Read the precautions on the back before filling this page) 414997 A7 B7 V. Description of the invention () Please refer to the second picture 'shown in this picture Control system for production 1 〇〇The manufacturing process flow chart of the manufacturing process 110. In complex semiconductor process steps, inspection stations are usually set up to ensure the quality of the process. Blocks 220, 230, 260, and 270 in the figure are the process steps. 〖'2, 3, and 4, when the crystal 0 passes through the execution of process step ^ 2 (blocks 220, 230)', it will enter block 240 to determine whether to perform the detection step, and the criterion for this judgment is the above. The sampling frequency of this product obtained by the background server 400. After the judgment, if the conditions do not meet, then skip the detection step 250 and execute the process steps 3 and 4, otherwise, it will enter the eight detection step called 25 0 to perform the inspection, and then Perform process steps 3 and 4. Reference is now made to the third diagram, which shows a schematic diagram of adjusting the frequency of product inspection in accordance with the process stability of the present invention. It has been described above that the system of the present invention can automatically adjust the detection sampling frequency according to process stability. When new products are introduced into production or the process stability is extremely low, the sampling frequency will increase (up to 1, which is 100% inspection), and After the process stability is increased, the process statistical analyzer 200 can learn from statistical calculations, and then store the lower sampling frequency in the sampling frequency database 300 for the background server 400 to access. The leftmost column in this figure is the process stability, the rightmost column is the detection sampling frequency, and the box shown in the middle is the process to be tested. The paper size is applicable to the National Standard (CNS) A4 specification (210x297). (Mm) (Please read the notes on the back before filling out this page) ^ 一 ^ ------ Order--I I --- 1 1 ^. Printed by the Consumers ’Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs A7 414997 __ B7__ 5. Description of the invention (). When the process stability before this test station is very low or newly introduced products (the first row in the figure) 'In order to find process problems to maintain manufacturing quality, the sampling frequency will be set to 1, at this time all products After this testing station will be tested (indicated by the area of the oblique line); when the process stability gradually stabilizes (the middle state in the second row in the figure), the sampling frequency is reduced to 1/2, which is two consecutive Only one of the manufacturing lot (lot) is selected for inspection; after the process stability becomes higher (the third row in the figure), the sampling frequency is further reduced to 1/3, which is the average of the three consecutive manufacturing lots. Only one of them is selected for inspection; however, if the process stability drops to a medium state (the fourth row in the figure), the sampling frequency will return to the state of ½ again. "After the above detailed description of the present invention, The sampling frequency of quality control in the manufacturing process can be controlled by statistical calculation methods through product maturity and stability. At the same time, “this system can not only manage the execution of the detection steps automatically in a systematic way” to save costs, but also can adjust the sampling frequency according to the actual process conditions through dynamic input. As understood by those familiar with this technology, the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the patent application for the present invention; all others completed without departing from the spirit disclosed by the present invention, etc. Effective changes or modifications should be included in the scope of patent application described below. Each paper size applies Chinese National Standard (CNS) A4 specification (210 X 297 public love) · lit -------- t --------- 0 I (Please read the precautions on the back before (Fill in this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs