What is considered a Parametric Outlier?
Test Failures - gross outlier removal
Parametric Outliers - atypical results in Bin 1 or good Bins population
Parametric Outlier Detection
Part Average Testing (PAT), as specified by the Automotive Electronics Council (AEC-Q001-REV C), involves DPPM improvement methodologies that address the identification of device parametric outliers in ”normal,” or Gaussian, data distributions. However, normal/Gaussian data distributions correspond to only a subset of typical device tests. Therefore in non-Gaussian data distributions, PAT-only outlier detection methods could either induce yield drops or potentially skip the detection of outliers.
Streetwise™ offers a comprehensive solution for outlier detection in BOTH Gaussian and non-Gaussian data distributions. The patented and patent-pending Streetwise™ technology is adaptive, and determines the data distribution on a per wafer and per test basis, taking in consideration test limit dependencies.
- Dynamic PAT: parametric outlier detection in “normal” Gaussian distribution.
|
 |
- Parametric outlier detection beyond traditional PAT and across all potential data population distributions using automatic algorithm selection based on the data population distribution encountered for a/any/all given test(s).
|
|
Streetwise™ Parametric Outlier Classification
Streetwise™ introduces advanced, recipe-driven classification of test outliers. This functionality fine-tunes outlier evaluation and ensures robust and reliable selection of outlier devices. Based on the user-defined recipe, Streetwise™ will automatically designate each class of outlier parts for either dispositioning or removal due to quality risk.
Test Failures - gross outlier removal
Parametric Outliers - atypical results in Bin 1 or good Bins population
Streetwise™ Adaptive Outlier Detection
Streetwise™ is:
- Adaptive: Streetwise™ automatically selects per-wafer and per-test the appropriate outlier detection algorithm from its library as required by the data population distribution and the test limits
- Automatic: no engineering data analysis required for set-up - ready to go “out-of-the-box” for production
- Configurable: allows customer to apply existing quality standards using the Streetwise™ recipe editor
