Gage R and R Studies for PCB Measurement System Analysis

Before a process can be controlled, it has to be measured, and the measurement itself has to be trustworthy. A thickness gauge, a microscope scale or a tester that reads inconsistently will produce data that looks like process variation. A Gage R and R study separates the variation coming from the measurement system from the variation coming from the process.

Why Measurement Systems Need Study

Every measurement contains error. The gauge, the operator, the fixture, the environment and the part itself all contribute, and the total is the variation seen in the readings. If that variation is large compared with the tolerance being checked, the measurements cannot distinguish a good part from a marginal one.

The problem is invisible in normal production because the numbers look plausible. Two operators measure the same panel and get different results, and the difference is attributed to the process. A study makes that ambiguity visible and quantifies it, so the decision about whether to trust the data is based on evidence.

Variation, Accuracy and Precision

Accuracy is how close the average reading is to the true value, and precision is how closely repeated readings agree with each other. A gauge can be precise without being accurate, which happens when it is consistently offset by a calibration error, and it can be accurate on average while scattering widely.

Both matter, but for process control the precision is usually the more important term, because a systematic offset can be corrected while random scatter cannot. A gauge that reads one hundredth of a millimetre high consistently is easier to live with than one that scatters over that range. Linearity, which describes how the offset changes across the measuring range, is a related property and should be checked whenever a gauge is used over a wide span.

The Gage R and R Method

The standard study uses several operators, several parts and several repeated measurements of each part by each operator. The parts are chosen to span the range of the process, not to be identical, because the study needs real part-to-part variation to compare against. The measurement order is randomised so that the operators do not remember previous readings.

The data are analysed to separate the variation between repeated measurements of the same part by the same operator, the variation between operators measuring the same part, and the variation between parts. The first two are the measurement system; the third is the process.

Repeatability and Reproducibility Explained

Repeatability is the variation seen when one operator measures the same part several times with the same instrument. It captures the gauge itself, the fixture and the inherent difficulty of the measurement. A measurement that requires judgement, such as reading a microscope scale, has poor repeatability by nature.

Reproducibility is the variation between operators. It captures differences in technique, in how the part is seated, in how the scale is read. Training reduces reproducibility error, and a fixture or a more automatic gauge can reduce it further by removing judgement from the measurement.

Setting Up a Study on the Shop Floor

A study needs parts, operators and time, and it will disrupt production if it is not planned. Ten parts spanning the range, two or three operators and two or three repeats each is the common arrangement, giving enough data for a useful estimate without taking the line out of service for long.

The parts should be labelled so that they cannot be identified during the study, and the operators should measure them in a random order without seeing previous results. If the operators can see the numbers, they will unconsciously adjust their technique, and the reproducibility result will be optimistic.

Reading the Results

The output is normally expressed as the measurement system variation as a percentage of the total variation, or as a percentage of the tolerance. A common interpretation is that below ten percent is acceptable, between ten and thirty percent is conditionally acceptable depending on the application, and above thirty percent needs improvement. The threshold should be chosen with the application in mind, because a measurement used for a final acceptance decision needs to be more capable than one used only for a rough process check.

The number should be read together with the breakdown. If reproducibility dominates, the issue is training or technique. If repeatability dominates, the issue is the gauge or the fixture. Those two conclusions lead to very different corrective actions, which is why the split matters as much as the total.

Measurement Systems in PCB Fabrication

Plating thickness is a classic case. A gauge that measures on a coupon and one that measures on a finished trace can disagree, and the difference matters when a specification is being checked. Establishing which method is authoritative, and how consistent each is, prevents disputes later, and a test coupon in the panel border is usually the practical way to make the measurement repeatable.

Dimension and registration measurements have the same issue, and so do the visual judgements made at final inspection. Where a decision is made by eye, a study of agreement between inspectors is the equivalent of a Gage R and R, and it usually shows wider variation than expected. The inspection methods involved are described in this guide to judging PCB quality.

Actions When the System Fails

An unacceptable result usually has a specific cause. A worn gauge, a fixture that does not seat the part repeatably, an unclear procedure, or an operator who has not been trained on the specific measurement. Fixing the cause is often cheap, and it improves every process that depends on the same measurement.

Where the measurement is inherently difficult, the answer may be to change it: measure a coupon instead of the product, use a fixture to locate the part consistently, or replace a manual gauge with one that removes judgement. Thickness measurement is a good example, and the options are described in this guide to plating thickness measurement.

Sustaining Measurement Capability

A study is a snapshot, so it should be repeated after a change of gauge, a change of procedure, a change of operator group or a significant repair. It should also be repeated periodically, because a gauge drifts and a fixture wears, and the drift is gradual enough to go unnoticed.

The results belong with the calibration record and the process documentation, so that the measurement basis for a specification is visible to anyone reviewing the data. Where a customer requires a measurement to be made a particular way, that method is what should be qualified, and the requirement should be written into the fabrication notes rather than agreed verbally.

FAQ

How many operators and parts does a study need? A common arrangement is three operators, ten parts spanning the process range and three repeats each. Fewer operators reduce confidence in the reproducibility estimate, and parts that do not span the range make the measurement system look worse than it is.

What if my measurement is destructive? Destructive tests cannot be repeated on the same part, so the study has to use parts that are assumed equivalent, such as adjacent coupons from the same panel. The repeatability estimate is then a combination of gauge and part variation, and it should be interpreted with that limitation in mind.

Is a Gage R and R the same as calibration? No. Calibration confirms that the gauge reads correctly against a standard, while a study examines the whole measurement system including operators, fixtures and procedure. A calibrated gauge can still be part of a measurement system that is not capable.