Methods and apparatus for local outlier detection

• US 8,417,477 B2
• Filed: 05/20/2005
• Issued: 04/09/2013
• Est. Priority Date: 05/24/2001
• Status: Expired due to Term

• Current Assignee: In-Depth Test LLC (Acacia Research Corporation)
• Original Assignee: Test Acuity Solutions, Inc (Boston Semi Equipment, LLC)
• Inventors: Miguelanez, Emilio, Gorin, Jacky, Tabor, Eric Paul
• Primary Examiner(s): BHAT, ADITYA S
• Application Number:
  US11/134,843
• Publication Number:
  US 20060085155A1
• Time in Patent Office: 2,881 Days
• Field of Search: 702/118
• US Class Current: 702/118
• CPC Class Codes: G01R 31/2836 Fault-finding or characterising G01R31/2822 - G01R31/2831 take precedence
• CPC Class Codes: G01R 31/2894 Aspects of quality control [QC] G01R31/31718 takes precedence; program control for QC G05B19/41875
• CPC Class Codes: G01R 31/31718 Logistic aspects, e.g. binning, selection, sorting of devices under test, tester/handler interaction networks, Test management software, e.g. software for test statistics or test evaluation, yield analysis mechanical aspects G01R31/2808, G01R31/2851
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04

Abstract

A method and apparatus for data analysis according to various aspects of the present invention is configured to identify statistical outliers in test data for components, including local outliers representing outliers within subsets of larger data populations.

Claims

1. A data analysis system for semiconductor test data, comprising a computer system, wherein the computer system is configured to operate:

a local outlier identification system configured to;

select a data subset from the semiconductor test data corresponding to a spatially-related subgroup of components on a substrate, wherein the spatially-related subgroup of components comprises a group of adjacent components within an area defined on the substrate that is fewer than all of the components on the substrate; and

automatically identify a local outliers in the data subset, wherein the local outlier comprises an outlier relative to the selected data subset.

2. The test system of claim 1, wherein the local outlier identification system is configured to normalize the test data.

3. The test system of claim 1, wherein the local outlier identification system is configured to identify the local outliers according to a local median and a local interquartile range of the data subset.

4. The test system of claim 1, wherein the local outlier identification system is configured to select the data subset according to a spatial pattern.

5. The test system of claim 4, wherein the local outlier identification system is configured to automatically adjust at least one of a size and a shape of the spatial pattern.

6. The test system of claim 4, wherein:

the spatial pattern defines a central component; and

the local outlier identification system is configured to identify the local outliers in the data subset by comparing the test data for the central component to a threshold value derived from a median value for the test data for the non-central components in the data subset.

7. The test system of claim 1, wherein the local outlier identification system is configured to automatically select an outlier identification algorithm for identifying the local outliers from a plurality of candidate outlier identification algorithms.

8. The test system of claim 1, wherein the local outlier identification system is configured to assign a category to a selected component according to at least one of a magnitude and a number of at least one of one or more global outliers and one or more local outliers identified in the test data for the component.

9. A data analysis system for semiconductor test data, comprising a computer system, wherein the computer system is configured to operate:

a local outlier identification system configured to;

select test data for a spatially-related subgroup of components on a selected wafer, wherein the spatially-related subgroup of components comprises a group of adjacent components within an area defined on the wafer that is fewer than all of the components on the selected wafer; and

identify local outliers in the selected test data, wherein the local outliers comprise an outlier relative to the selected test data.

10. The test system of claim 9, wherein the local outlier identification system is configured to normalize the test data.

11. The test system of claim 9, wherein the local outlier identification system is configured to identify the local outliers according to a local median and a local interquartile range of the selected data.

12. The test system of claim 9, wherein the local outlier identification system is configured to select the selected data according to a spatial pattern.

13. The test system of claim 12, wherein the local outlier identification system is configured to automatically adjust at least one of a size and a shape of the spatial pattern.

14. The test system of claim 12, wherein:

the spatial pattern defines a central component; and

the local outlier identification system is configured to identify the local outliers in the selected data by comparing the test data for the central component to a threshold value derived from a median value for the test data for the non-central components in the selected data.

15. The test system of claim 9, wherein the local outlier identification system is configured to automatically select an outlier identification algorithm for identifying the local outliers from a plurality of candidate outlier identification algorithms.

16. The test system of claim 9, wherein the local outlier identification system is configured to assign a category to a selected component according to at least one of a magnitude and a number of at least one of one or more global outliers and one or more local outliers identified in the test data for the component.

17. A test data analysis system for analyzing test data for a set of components fabricated and tested using a fabrication process, comprising:

a memory for storing the test data; and

a local outlier identification system having access to the memory and configured to;

select a portion of the test data corresponding to a group of spatially-related components on a single substrate comprising a group of adjacent components within a defined area on the single substrate that is fewer than all of the components on the substrate; and

identify local outliers in the selected portion, wherein the local outliers comprise an outlier relative to the selected portion of the test data.

18. The test data analysis system of claim 17, wherein the local outlier identification system is configured to normalize the test data.

19. The test data analysis system of claim 17, wherein the local outlier identification system is configured to identify the local outliers according to a local median and a local interquartile range of the portion of the test data.

20. The test data analysis system of claim 17, wherein the local outlier identification system is configured to select the portion of the test data according to a spatial pattern.

21. The test data analysis system of claim 20, wherein the local outlier identification system is configured to automatically adjust at least one of a size and a shape of the spatial pattern.

22. The test data analysis system of claim 20, wherein:

the spatial pattern defines a central component; and

the local outlier identification system is configured to identify the local outliers in the portion of the test data by comparing the test data for the central component to a threshold value derived from a median value for the test data for the non-central components in the portion of the test data.

23. The test data analysis system of claim 17, wherein the local outlier identification system is configured to automatically select an outlier identification algorithm for identifying the local outliers from a plurality of candidate outlier identification algorithms.

24. The test data analysis system of claim 17, wherein the local outlier identification system is configured to assign a category to a selected component according to at least one of a magnitude and a number of at least one of one or more global outliers and one or more local outliers identified in the test data for the component.

25. A computer-implemented method for testing components on a selected substrate fabricated and tested according to a fabrication process, comprising:

obtaining test data for the components;

selecting a subset of the test data corresponding to a spatially-related subgroup of components on the selected substrate, wherein the subgroup comprises a group of adjacent components within a defined area on the substrate that is fewer than all of the components on the substrate;

automatically identifying local outliers in the selected subset of test data, wherein the local outliers comprise an outlier relative to the selected subset of the test data; and

storing information identifying the outliers in a computer readable medium.

26. The computer-implemented method of claim 25, further comprising normalizing the test data.

27. The computer-implemented method of claim 25, wherein identifying the local outliers comprises determining a local outlier threshold according to a local median and a local interquartile range of the data subset.

28. The computer-implemented method of claim 25, wherein selecting the subset comprises selecting the data subset according to a spatial pattern.

29. The computer-implemented method of claim 28, wherein selecting the subset further comprises automatically adjusting at least one of a size and a shape of the spatial pattern.

30. The computer-implemented method of claim 28, wherein:

the spatial pattern defines a central component; and

identifying the local outliers in the data subset comprises comparing the test data for the central component to a threshold value derived from a median value for the test data for the non-central components in the data subset.

31. The computer-implemented method of claim 25, further comprising automatically selecting an outlier identification algorithm for identifying the local outliers from a plurality of candidate outlier identification algorithms.

32. The computer-implemented method of claim 25, further comprising assigning a category to a selected component according to at least one of a magnitude and a number of at least one of one or more global outliers and one or more local outliers identified in the test data for the component.

33. A non-transitory computer-readable medium storing instructions executable by a computer, wherein the instructions cause the computer to execute a method for analyzing test data comprising:

obtaining test data for the components on a selected substrate;

selecting a portion of the test data corresponding to geographically-related components on the selected substrate, wherein the selected portion comprises a group of adjacent components within a defined area on the substrate that corresponds to fewer than all of the components on the selected substrate; and

identifying a local outlier in the selected portion of the test data, wherein the local outliers comprise an outlier relative to the selected portion of the test data.

34. The computer-readable medium storing instructions of claim 33, the method further comprising the step of normalizing the test data.

35. The computer-readable medium storing instructions of claim 33, wherein identifying the local outlier comprises determining a local outlier threshold according to a local median and a local interquartile range of the portion of the test data.

36. The computer-readable medium storing instructions of claim 33, wherein selecting the portion of the test data comprises selecting the portion of the test data according to a spatial pattern.

37. The computer-readable medium storing instructions of claim 36, wherein selecting the portion of the test data further comprises automatically adjusting at least one of a size and a shape of the spatial pattern.

38. The computer-readable medium storing instructions of claim 36, wherein:

the spatial pattern defines a central component; and

identifying the local outlier in the portion of the test data comprises comparing the test data for the central component to a threshold value derived from a median value for the test data for the non-central components in the portion of the test data.

39. The computer-readable medium storing instructions of claim 33, the method further comprising automatically selecting an outlier identification algorithm for identifying the local outlier from a plurality of candidate outlier identification algorithms.

40. The computer-readable medium storing instructions of claim 33, the method further comprising assigning a category to a selected component according to at least one of a magnitude and a number of at least one of one or more global outliers and one or more local outliers identified in the test data for the component.