Semi E89

SEMI E89 defines Measurement System Analysis (MSA) guidelines for semiconductor metrology — the statistical framework for quantifying how much of a measurement's observed variation comes from the measurement system itself (gauge repeatability, reproducibility, linearity, bias) versus the actual part-to-part variation the measurement is supposed to be detecting. A metrology tool that reports a wafer's film thickness or critical dimension isn't just reporting the true value — it's reporting the true value plus whatever noise and systematic error the measurement system itself contributes, and E89 is the standard that gives fabs a consistent way to quantify that contribution before trusting the tool's numbers for process control decisions.

Metrology tool reports a measured value
Film thickness, CD, overlay — any process-control measurement
SEMI E89 — Measurement System AnalysisSeparating measurement noise from real part-to-part variation
Repeatability and reproducibility (Gauge R&R)
Same tool/operator variation vs. tool-to-tool, operator-to-operator
Linearity and bias across the measurement range
Systematic error that isn't random noise
Fab knows how much of the reported variation is real
Process-control decisions aren't chasing measurement noise

The reason this matters beyond a one-time tool qualification is that every downstream process-control decision — SPC control limits, R2R recipe adjustments, yield correlation — inherits whatever uncertainty the measurement system contributes, and if that uncertainty isn't characterized, a fab risks either tightening process control around measurement noise that isn't real variation, or missing genuine process drift that's smaller than the measurement system's own error. E89's MSA framework is what lets a fab draw the line between "this measurement moved because the process moved" and "this measurement moved because the gauge itself is noisy," which is the precondition for every other metrology-dependent standard in the SEMI catalog — virtual metrology, R2R control, SPC — to actually mean what it claims to mean.

SEMI E89 — Splitting Observed Variation measurement noise vs. real process variation Total observed variation in a measured value Measurement-system variation Gauge R&R, linearity, bias Should be minimized, not chased Real part-to-part variation What SPC/R2R should respond to The signal that actually matters Confusing the two leads to chasing noise or missing real drift

E89 is foundational rather than peripheral: any metrology-dependent control scheme is only as trustworthy as the measurement system analysis behind it, which is why E89 functions as the prerequisite statistical discipline underneath virtual metrology, R2R control, and every other SPC-based standard in the fab's process-control stack.

Take SEMI E89 further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app