Gauge R&r
Gauge R&R (repeatability and reproducibility) is the statistical study that answers a question a fab has to settle before it trusts any measurement at all: how much of the variation a tool reports is the part actually varying, and how much is just noise from the measurement system itself.
Gauge R&R solves the problem that every measurement includes some amount of noise from the act of measuring itself, and without separating that noise from real part-to-part variation, a fab can't tell whether a number it's looking at describes the wafer or just describes the measurement. A metrology tool reporting slightly different critical-dimension values on the same die measured twice isn't necessarily seeing real variation — some of that spread is the tool's own repeatability limit, and some can come from which operator loaded the wafer or which day the measurement was taken — Gauge R&R exists specifically to split total observed variation into those pieces so a fab knows how much of what it's measuring is signal and how much is just the cost of measuring at all.
The result comes out as a single percentage, and that number has industry-standard thresholds a fab actually acts on. Gauge R&R is typically expressed as a percentage of total variation that the measurement system itself accounts for: below 10 percent is generally considered an acceptable measurement system, 10 to 30 percent is marginal and may be acceptable depending on the application's cost and criticality, and above 30 percent means the measurement system's own noise is too large relative to the real part-to-part variation to trust for process control — a tool that fails Gauge R&R isn't necessarily broken, but its numbers can't be trusted to drive a yield or process decision until the noise source is found and reduced.
| Gauge R&R result | What it means | Typical action |
|---|---|---|
| Below 10% | Measurement system is acceptable | Trust the tool for process control |
| 10% to 30% | Marginal, judgment call | May be acceptable for lower-criticality measurements |
| Above 30% | Measurement system is unacceptable | Investigate tool, operator procedure, or fixture before trusting the data |
A failed Gauge R&R study points investigation in a specific direction depending on which component of variation turned out to be large. A high repeatability component — the same operator, same tool, repeated measurement spread — points at the tool itself: wear, drift, poor fixturing, or a measurement technique that's inherently noisy at the precision being demanded. A high reproducibility component — different operators getting different answers on the same part — points instead at operator technique or an ambiguous measurement procedure, which is a training and documentation problem rather than a hardware one; separating the two is exactly why Gauge R&R studies deliberately vary both tool and operator rather than just repeating one measurement over and over.
Gauge R&R sits inside the broader discipline of measurement system analysis as the specific statistical technique that quantifies repeatability and reproducibility, rather than being a synonym for the whole field. A fab typically runs a Gauge R&R study before trusting a new metrology tool with production decisions, after any major tool service or recalibration, and periodically on critical measurement systems to confirm nothing has quietly drifted — it's one of the concrete, numeric tools a fab uses to answer the broader measurement system analysis question of whether a given tool and procedure are fit to make the decision they're being asked to make.
Read Gauge R&R through a signal-versus-noise lens: the number coming off a metrology tool was never guaranteed to be pure signal, and Gauge R&R is the structured way a fab finds out how much of what it's measuring is the wafer and how much is just the unavoidable noise of measuring it — a distinction that has to be settled before any of that data is trusted to drive a real process decision.