PCB Measurement System Analysis Guide

Every number a board shop uses to control a process comes from a measurement, and every measurement carries its own variation. Measurement system analysis is the work of finding out how much of the variation in a set of readings comes from the gauge and the person using it, rather than from the process being measured. It matters because a gauge that contributes a third of the observed spread makes a capable process look incapable, and sends improvement effort in the wrong direction for months.

Why Measurement Variation Matters

A process is judged entirely from measurements, so the quality of the measurement sets the limit on what can be known about the process. If a plating thickness reading varies by two micrometres when the same coupon is measured twice, then a difference of one micrometre between two batches cannot be interpreted at all. Teams that do not know this will chase differences that are entirely inside the noise of their own gauge, and they will do it repeatedly because no result ever looks convincing.

The effect is worse when a specification limit is close to the measurement spread. A process with a real capability of one and a third can be reported as having a capability below one if the gauge contributes a significant share of the observed variation, and the shop will then spend money on the process when the money was needed for a better measurement. Separating the two is therefore not an academic exercise but a direct input to where improvement effort should go.

Repeatability and Reproducibility

Repeatability is the variation seen when one person measures the same item several times with the same gauge, and it reflects the equipment and the method. Reproducibility is the variation seen when different people measure the same item, and it reflects the training, the technique and the interpretation of the specification. A gauge r&r study measures both together, and the proportion each contributes tells the shop whether to invest in equipment or in instruction.

A simple study uses two or three operators, ten parts that span the expected range of the measurement, and two or three repeats per operator. The arithmetic produces a repeatability figure, a reproducibility figure and a combined value expressed as a percentage of the total variation or of the tolerance. A rule of thumb is that a combined figure under ten percent is good, between ten and thirty percent is acceptable depending on the application, and above thirty percent needs attention before the measurement is used for control.

Running the Study in a Board Shop

The practical difficulty in a board shop is finding parts that span the range without them being obviously different. Ten coupons from the same panel may all be within a micrometre of each other, which will produce a study that says nothing. The samples should be selected deliberately across the tolerance range, and where the process is very consistent, samples from the limits of the tolerance may have to be made specially.

Operators in the study should not know which coupon is which, and they should not see each other’s readings. Both precautions remove the tendency to compare and revise, which is a normal human behaviour and which destroys the study. The measurement should also follow the normal method exactly, including the number of readings and the position on the coupon, because a study performed under laboratory conditions does not describe what happens in production.

Calibration Is Not Enough

Calibration confirms that a gauge agrees with a reference at the time of calibration, and it says nothing about how a person uses the gauge from day to day. A calibrated micrometer used at a different point on the barrel, or closed with a different feel, will produce variation that no certificate can detect. This is the most common misunderstanding in measurement control, and it is why gauge r&r and calibration are complementary rather than alternatives.

The measurement method matters as much as the instrument. Where on the coupon the plating thickness is measured, how many points are averaged, whether the probe is allowed to settle and how the reading is rounded all belong in the method, and all of them should be written down. A method that exists only in the habits of the most experienced operator cannot be taught, audited or improved, and it will produce different results on different shifts.

Using the Result

The result of a measurement system analysis should change something. If reproducibility is poor, the fix is usually a written method and a short training session on the gauge. If repeatability is poor, the gauge may be unsuitable, worn or being used outside its range. If the study is acceptable, the shop has a measurement it can trust for process control, and that trust is what makes a control chart meaningful rather than decorative.

The study should be repeated when the gauge, the method or the operators change, and periodically even when they do not, because equipment drifts and habits change. Recording the studies with the fabrication records keeps them available when a capability figure is questioned, and it demonstrates to a customer that the numbers behind a report have been tested rather than assumed.

Measurement and Process Control

A control chart is only as good as the measurement underneath it. If the gauge contributes a large share of the variation, the chart will show a wide band of noise and will fail to detect a genuine drift until it has become a large one. This is the mechanism by which a shop with good gauges detects a problem a week earlier than a shop with poor ones, without any difference in the processes themselves.

The checking measurements are particularly important. Plating thickness is usually measured on a coupon rather than on the board, and the plating specification defines where and how that measurement should be taken. If the coupon measurement is imprecise, the entire control loop is imprecise, and the process will be adjusted on the basis of noise. Measurement system analysis on the coupon method is therefore one of the highest value studies a shop can perform.

Practical Rules

Study the measurement before drawing conclusions from the numbers, choose samples that span the range, and keep operators from comparing readings. Separate repeatability from reproducibility so that the fix can be aimed at the equipment or at the method.

Write the method down, repeat the study when anything changes, and apply it to the measurements used for quality decisions as well as to those used for process control. A shop that knows how good its measurements are can act on smaller differences, and that ability is worth more than any single improvement project, because it applies to every decision that follows.

FAQ

What does a gauge r&r study measure? Repeatability, which is the variation of one operator with one gauge, and reproducibility, which is the variation between operators. Together they show how much of the spread comes from the measurement.

How many samples are needed? Typically ten parts spanning the expected range, measured by two or three operators with two or three repeats each.

Is calibration sufficient? No. Calibration shows the gauge agrees with a reference. It says nothing about how the gauge is used, which is what a gauge r&r study examines.