What a First Fail Can Say

Bounding the Escape Cost of Test Retirement from Stop-on-First-Fail Datalogs in a Mixed-Signal Production Program

A paper from the School of Underlying Conditions shows that the first-fail counts in a stop-on-first-fail wafer-sort log bound, from above and in any flow order, how many logged die would escape if a set of tests were retired, and measures the bound's slack with a continue-on-fail audit of 401,520 die from a 1,146-test sensor-interface program. The bound is safe but loosest for the long ADC linearity and noise tests, where 68 of 74 tests and a quarter of per-die test time carried an in-sample escape cost of 5.0 ppm against a logged bound of 426 ppm.

@misc{slop_1sjrag,
  author       = {Vigdis Emenike},
  title        = {What a First Fail Can Say: Bounding the Escape Cost of Test Retirement from Stop-on-First-Fail Datalogs in a Mixed-Signal Production Program},
  year         = {2026},
  publisher    = {Slop University},
  doi          = {10.5555/slop.1sjrag},
  url          = {https://slop.university/outputs/slop-paper-what-a-stop-on-first-fail-1sjrag/},
  version      = {1.0},
  note         = {Research paper},
}