Method of statistical binning for reliability selection

• US 6,789,032 B2
• Filed: 12/19/2002
• Issued: 09/07/2004
• Est. Priority Date: 12/26/2001
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Barnett, Thomas S., Purdy, Kathleen G., Grady, Matthew S., Barbour, Tange N.
• Primary Examiner(s): SHAH, KAMINI S
• Application Number:
  US10/326,668
• Publication Number:
  US 20030120445A1
• Time in Patent Office: 628 Days
• Field of Search: 702/81-84, 702/117, 702/118, 702/181, 702/33, 702/34, 702/179, 702/182, 700/109, 700/11, 700/121
• US Class Current: 702/81
• 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 statistical method is described for reliability selection of dies on semiconductor wafers using critical wafer yield parameters. This is combined with other data from the wafer or module level reliability screens (such as voltage screen or burn-in) to obtain the relative latent defect density. Finally the modeled results are compared with actual results to demonstrate confidence in the model.

Claims

1. A statistical method for reliability selection of dies on semiconductor wafers to be tested comprising:

collecting test data from each of a plurality of wafers to obtain actual results of each die;

extracting a yield model of certain parameters from the test data; and

comparing the yield model and actual reliability data to obtain a distribution of good die reliability fails based on negative binomial statistical modeled results and actual results.

2. The method of claim 1 which includes:

analyzing the test data to determine relationships between the dies.

3. The method of claim 2 in which the resultant relationship is the location of the failing dies.

4. The method of claim 3 in which the resultant relationship is used to determine the number of fails of neighboring die surrounding any given die.

5. A statistical method for reliability selection of dies on semiconductor wafers to be tested comprising:

collecting test data from each of a plurality of wafers to obtain actual results of each die;

extracting a yield model of certain parameters from the test data;

analyzing the test data to determine relationships between the dies wherein the resultant relationship is the number of embedded repairs on a particular die; and

comparing the yield model and actual reliability data to obtain a distribution of good die reliability fails based on modeled results and actual results.

6. A statistical method for reliability selection of dies on semiconductor wafers to be tested comprising:

collecting test data from each of a plurality of wafers to obtain actual results of each die;

extracting a yield model of certain parameters from the test data wherein parameters include killer defects per wafer, degree of clustering of idler defects, and the ratio of latent defects to killer defects; and

comparing the yield model and actual reliability data to obtain a distribution at good die reliability fails based on modeled results and actual results.

7. A statistical method for reliability selection of dies on semiconductor wafers to be tested comprising:

collecting test data from each of a plurality of wafers to obtain actual results of each die;

comparing the yield model and actual reliability data to obtain a distribution of good die reliability fails based on modeled results and actual results wherein the parameters are used to calculate the distribution of good die in each bin category and distribution of reliability fails in each bin.

8. A statistical method for reliability selection of dies on semiconductor wafers to be tested comprising:

collecting test data from each of a plurality of wafers,

determining certain parameters based on average number of killer defects in a wafer;

computing the number of embedded repairs for each die;

classifying the die into a number of bins based on the number of repairs and clustering to determine the distribution of good die in each bin; and

comparing the classification of the die with actual distributions.

9. A statistical method for reliability selection of dies on semiconductor wafers to be tested comprising:

collecting test data from each of a plurality of wafers;

determining certain parameters based on average number of killer defects in a wafer wherein the parameters include clustering of defects and defect density;

computing the number of embedded repairs for each die;

classifying the die into a number of bins based on the number of repairs and clustering to determine the distribution of good die in each bin; and

comparing the classification of the die with actual distributions.

10. The method of claim 9 wherein the die in each bin will have a different probability of possessing a reliability defect.

11. The method of claim 10 wherein the bin with zero repairs will have the lowest probability of having a reliability defect.