Confidence Interval and Number of Samples

The change in confidence interval resulting from an increase or decrease in the total number of samples cannot be determined in advance because it depends on the given circuit and failure boundary.

You can generate a quantitative example by running the Scales-Sigma Sampling method on a linear failure boundary case.

The following table reports the 5% and 95% quantile values generated based on 100 Scales-Sigma Sampling runs for each result (combination of number of samples and sigma target).

Number of Samples	4 Sigma	6 Sigma
1000	3.30 4.75	4.56 7.49
7000	3.62 4.32	5.31 6.75
50000	3.80 4.12	5.70 6.36

Depending on the failure boundary shape, the confidence interval may differ. The linear boundary results can be used as a reference because of the following reasons:

• Many failure boundaries are not very strongly nonlinear.
• The confidence interval tends to be:
  • Small: If the failure rate scales faster than linear boundary.
  • Large: If failure rate scales slower than linear boundary.
    
    

Related Topics

High Yield Estimation

The Scaled-Sigma Sampling Method

Running the Scaled-Sigma Sampling Method

Return to top