Method for analyzing in-line QC test parameters

• US 6,959,252 B2
• Filed: 07/03/2003
• Issued: 10/25/2005
• Est. Priority Date: 12/31/2002
• Status: Expired due to Fees

• Current Assignee: Powerchip Semiconductor Corporation (Powerchip Technology Corporation)
• Original Assignee: Powerchip Semiconductor Corporation (Powerchip Technology Corporation)
• Inventors: Tai, Hung-En, Luo, Haw-Jyue
• Primary Examiner(s): Barlow, John
• Assistant Examiner(s): Le, Toan M.
• Application Number:
  US10/604,244
• Publication Number:
  US 20050004773A1
• Time in Patent Office: 845 Days
• Field of Search: 702/84, 700/109, 438/14
• US Class Current: 702/84
• CPC Class Codes: G05B 19/41875 characterised by quality surveillance of production
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

A method for analyzing in-line QCtest parameters is used to analyze a plurality of lots of products, each lot of products having a lot number and being formed using a plurality of equipments. At least one wafer of each lot of products is tested by at least one in-line QC test item to generate an in-line QC test parameter. The in-line QC test item, a sample test item and a wafer test item related to the in-line QC test item are stored in a database. The database further stores the in-line QC test parameter and data of a plurality of lots of high-yield product stocks, such as various test items and test parameters. The method includes the following steps: analyzing the in-line QC test parameter to determine whether the in-line QC test parameter corresponds to a predetermined spec or not; searching the database to find out the sample test item or the wafer test item related to the in-line QC test item when the in-line QC test parameter does not correspond to the predetermined spec; searching the database to find out the corresponding test parameters of the high-yield product stocks according to the in-line QC test item and the searched sample test item or the wafer test item; and generating a correlation to illustrate the relationship between the in-line QC test item and the sample test item, or the relationship between the in-line QC test item and the wafer test item according to the searched high-yield product stocks.

Claims

1. A method for analyzing in-line quality control (QC) test parameters, the method being used to analyze a plurality of lots of products, each lot of products comprising a lot number, the products being formed using a plurality of equipments, at least one wafer of each lot of products being tested by at least one in-line QC test item to generate an in-line QC test parameter, the in-line QC test item, and its related sample test item and wafer test item being stored in a database, the database further storing the in-line QC test parameter and data of a plurality of lots of high-yield product stocks, such as test items and test parameters, the method comprising:

analyzing the in-line QC test parameter of the wafer to determine whether the in-line QC test parameter corresponds to a predetermined spec or not;

searching the database to find out the sample test item or the wafer test item related to the in-line QC test item when the in-line QC test parameter of the wafer does not correspond to the predetermined spec;

searching the database to find out the corresponding test parameters in the sample test item or the wafer test item of the high-yield product stocks according to the in-line QC test item and the searched sample test item or the wafer test item; and

using the in-line QC test parameter of the wafer and the searched test parameters in the sample test item of the high-yield product stocks to generate a correlation to illustrate the relationship between the in-line QC test item and the sample test item and using the in-line QC test parameter of the wafer and the correlation to predict a test parameter for the wafer in the sample test item, or using the in-line QC test parameter of the wafer and the searched test parameters in the wafer test item of the high-yield product stocks to generate a correction to illustrate the relationship between the in-line QC test item and the wafer test item and using the in-line QC test parameter of the wafer and the correlation to predict a test parameter for the wafer in the wafer test item.

2. The method of claim 1 wherein the lots of products are not yet tested by the sample test item and the wafer test item but obtain the predicted test parameter for each wafer of the lots of products in the sample test item or the wafer test item using the in-line QC test parameter of the wafer and the correlation between the in-line QC test item and the sample test item or the wafer test item.

3. The method of claim 1 wherein the correlation between the in-line QC test item and the sample test item, and the correlation between the in-line QC test item and the wafer test item are generated using linear regression methods.

4. The method of claim 1 wherein the database stores data of a process step related to the in-line QC test item, and the method further comprises:

classifying the lots of products into two groups according to a first spec, the two groups of products comprising a qualified group of products corresponding to the first spec, and a failed group of products not corresponding to the first spec;

searching the database to find out the process step related to the in-line QC test item;

finding the equipments used in the process step according to the lot numbers of the two groups of products; and

determining the equipment through which a probability that the failed group of products have passed is higher than a probability that the qualified group of products have passed.

5. The method of claim 4 wherein commonality analysis is used to determine the equipment through which a probability that a low-yield group of products have passed is higher than a probability that a high-yield group of products have passed.

6. The method of claim 1 further comprising:

searching test results of each of the sample test items and each of the in-line QC test items of the lots of products after a sample test process of the lots of products; and

generating a correlation between each of the sample test items and each of the in-line QC test items according to the searching results.

7. The method of claim 6 wherein the correlation between each of the sample test items and each of the in-line QC test items is generated by a multiple regression model.

8. The method of claim 6 wherein the correlation between each of the sample test items and each of the in-line QC test items is generated by a stepwise regression model.

9. The method of claim 6 wherein the correlation between each of the sample test items and each of the in-line QC test items is illustrated by a residual plot.

10. The method of claim 1 further comprising:

searching test parameters of each of the sample test items and each of the in-line QC test items of the lots of products when the lots of products are tested by the sample test item;

classifying the lots of products into a plurality of groups according to the parameters of each of the in-line QC test items of the lots of products;

analyzing the sample test parameters of each group of products; and

analyzing and obtaining the group of products having the sample test parameters most similar to a second spec when the sample test parameters of the groups of products are different.

11. The method of claim 10 wherein an ANOVA method is used to analyze whether the sample test parameters of the groups of products are different or not.

12. The method of claim 10 wherein a Duncan'"'"'s multiple range test is used to analyze and obtain the group of products having the sample test parameters most similar to the predetermined spec.

13. The method of claim 10 wherein the classified lots of products are illustrated by a box plot.

14. The method of claim 10 wherein each of the in-line QC test parameters of the obtained group of products is used as a predetermined spec of the subsequent products.

15. A method for analyzing in-line quality control test parameters, the method being used to analyze a plurality of lots of products, each lot of products comprising a lot number, the products being formed using a plurality of equipments, at least one wafer of each lot of products being tested by at least one in-line QC test item to generate an in-line QC test parameter, the in-line QC test item, and its related sample test item and wafer test item being stored in a database, the database further storing the in-line QC test parameter, data of a process step related to the in-line QC test item, and data of a plurality of lots of high-yield product stocks, such as test items and test parameters, the method comprising:

classifying the lots of products being tested by the in-line QC test item into two groups according to a first spec, the two groups of products comprising a qualified group of products which has the in-line QC test parameters corresponding to the first spec, and a failed group of products which has the in-line QC test parameters not corresponding to the first spec;

searching the database to find out the process step related to the in-line QC test item;

finding the equipments used in the process step according to the lot numbers of the two groups of products;

determining the equipment through which a probability that the failed group of products have passed is higher than a probability that the qualified group of products have passed, the equipment having the higher probability that the failed group of products have passed through being the possibly faulty equipment;

searching the database to find out the sample test item or the wafer test item related to the in-line QC test item;

searching the database to find out the corresponding test parameters in the sample test item or the wafer test item of the high-yield product stocks according to the in-line QC test item and the searched sample test item or the wafer test item; and

using the in-line QC test parameters of the failed group of products and the searched test parameters in the sample test item or the wafer test item of the high-yield product stocks to generate a correlation to illustrate the relationship between the in-line QC test item and the sample test item, or the relationship between the in-line QC test item and the wafer test item.

16. The method of claim 15 wherein the lots of products are not yet tested by the sample test item and the wafer test item and the correlation between the in-line QC test item and the sample test item or the wafer test item is utilized to predict test parameter for each wafer of the lots of products in the sample test item or the wafer test item.

17. The method of claim 15 wherein the correlation between the in-line QC test item and the sample test item, and the correlation between the in-line QC test item and the wafer test item arc generated using linear regression methods.

18. The method of claim 15 wherein commonality analysis is used to determine the equipment through which a probability that a low-yield group of products have passed is higher than a probability that a high-yield group of products have passed.