Measurement system analysis (MSA) is very important in the semiconductor industry to estimate the quality of the measurements. Most MSA indicators, especially the precision to tolerance (P/T) ratio, implicitly assume a normal distribution, with +/- kσ covering a given percentage of the distribution. In the reference documents (AIAG MSA Manual), there are no alternative calculations for non-normal data, and it is difficult to find a simple method that adapts to parameters with very different distributions.

We present two methods, with simple calculations and that are distribution agnostic, that cover the percentage of distribution set for our confidence level. The first method uses the Bienaymé-Tchebychev inequality to properly define the number of standard deviations in a k-sigma type formula. The second method uses a calculation of half-standard deviation on the right and on the left to allow for better coverage in the case of an asymmetric distribution. 

The two methods are applied on many electrical tests with JMP formulas and can generalize to outlier detection and removal.

Presenters

Schedule

Wednesday, 12 Mar
10:45-11:30

Room: Salon 5- London

Skill level

Intermediate
  • Beginner
  • Intermediate
  • Advanced

Files

Published on ‎12-15-2024 08:24 AM by Community Manager Community Manager | Updated on ‎02-03-2025 09:14 AM

Measurement system analysis (MSA) is very important in the semiconductor industry to estimate the quality of the measurements. Most MSA indicators, especially the precision to tolerance (P/T) ratio, implicitly assume a normal distribution, with +/- kσ covering a given percentage of the distribution. In the reference documents (AIAG MSA Manual), there are no alternative calculations for non-normal data, and it is difficult to find a simple method that adapts to parameters with very different distributions.

We present two methods, with simple calculations and that are distribution agnostic, that cover the percentage of distribution set for our confidence level. The first method uses the Bienaymé-Tchebychev inequality to properly define the number of standard deviations in a k-sigma type formula. The second method uses a calculation of half-standard deviation on the right and on the left to allow for better coverage in the case of an asymmetric distribution. 

The two methods are applied on many electrical tests with JMP formulas and can generalize to outlier detection and removal.



Will you be attending?

No one has RSVP'd 'Yes' yet

Add to Calendar
Starts:
Wed, Mar 12, 2025 05:45 AM EDT
Ends:
Wed, Mar 12, 2025 06:30 AM EDT
Salon 5- London
0 Kudos