Process Capability Cpk PCB Guide
Process capability is a statistical description of how well a process holds a dimension relative to its tolerance. The index that is usually quoted is Cpk, and it is a ratio between the distance from the process centre to the nearest limit and three standard deviations of the process. It is a useful number and an easily abused one, because it can be improved by widening a tolerance rather than by fixing a process.
What Capability Means
A capable process is one that produces parts inside the tolerance consistently, with a margin that absorbs the ordinary variation of the process.
The measurement is the distribution of a dimension over many parts, and the assumption is that the distribution is stable and roughly normal. Without stability the index describes a snapshot rather than a process.
The dimension has to be measurable and repeatable, which is a real constraint in a PCB shop. A hole diameter is easy to measure, while a solder fillet shape is not.
The value of the exercise is that it separates the variation that is built into the process from the variation that comes from something going wrong, and the two need different responses.
Cpk and Cp
Cp compares the width of the tolerance with the width of the process, and it ignores where the process is centred. A process that is off centre can have a good Cp and produce scrap.
Cpk accounts for the centring, and it is the smaller of the two distances from the mean to the limits, divided by three standard deviations. That is why Cpk is always less than or equal to Cp.
A common requirement is a Cpk of one and a third, which gives about four standard deviations between the mean and the nearest limit. A value above two is considered excellent and is usually achieved only by a very well controlled process.
The index says nothing about the shape of the distribution, so a process with two peaks can have an acceptable Cpk and still produce parts at both extremes.
<img src="https://www.gopcba.com/wp-content/uploads/2026/09/55-1.jpg" alt="Control chart showing a process distribution against limits” />
What the Index Needs
The first requirement is a stable process, which means the variation comes only from common causes. A control chart is the tool that establishes this, and a capability study on an unstable process is meaningless.
The second is enough data. A few dozen parts give a rough estimate, while a proper study uses a hundred or more measurements taken over a period that covers the ordinary sources of variation.
The third is a measurement system that is good enough. A gauge that contributes a large part of the observed variation will make a capable process look incapable, and the gauge should be studied before the process.
The fourth is a defined specification, since the index is calculated against the limits. A tolerance that is arbitrary makes the index arbitrary as well.
Applying It to a PCB Process
A hole diameter is a good candidate, because it is measured directly and it has a tolerance. The plating thickness is another, and it is measured on a coupon rather than on the board.
An impedance is a candidate if the measurement is repeatable, and the index is then used to check the stackup and the etching together.
The copper thickness, the board thickness and the outline dimensions are all used in the same way, and each one has its own distribution and its own tolerance.
A dimension that has no natural tolerance, such as the position of a component after placement, can still be studied, but the limits have to be set from the assembly requirement rather than from a drawing.
Control Charts
A control chart plots the measurement over time with a centre line and limits calculated from the process itself. It distinguishes a drift from a random fluctuation, which a capability number cannot do.
The chart is the operational tool and the capability index is the summary. A shop that calculates a Cpk once a year and never plots a chart is doing reporting rather than control.
A signal on the chart, such as a run of points on one side of the centre line, should trigger an investigation even though every point is inside the specification. That behaviour is what precedes a defect.
The chart also shows the effect of a change, such as a new drill or a new bath, which is how a process is improved rather than merely measured.
Limits and Misuse
The index can be improved by widening the tolerance, which is not an improvement at all. The specification should come from the design requirement rather than from the process result.
A capability study on a non normal distribution, such as a dimension that is bounded on one side, gives a misleading index unless a different method is used.
A high Cpk on a sample that was taken from one machine, one shift and one operator describes that sample and not the process. The sampling has to cover the variation that the product will see.
The number should be used to prioritise improvement rather than to certify a supplier, since a supplier with a good Cpk on one dimension can be incapable on another.
Practical Rules
Confirm that the process is stable before calculating a capability index, and use a control chart to do it.
Check the measurement system first, because a poor gauge makes a good process look bad.
Take the sample across machines, shifts and time so that the index describes the process rather than a moment.
Set the specification from the design requirement, and record the result with the fabrication notes and the quality data.
Process Control and Verification
On a design of this kind, process capability is the item that decides how the rest of the board is arranged. A stack-up that is drawn rather than described removes most of the ambiguity from a quotation, and it lets the fabricator price the board against the dielectric and copper weights that will actually be used. Where the requirement is not written down, the shop supplies its own default, and the default is chosen for the process rather than for the design.
The measurements that matter are the repeatable ones: conductor width and spacing, annular ring, finished hole size, plating thickness and surface finish are all verifiable on a coupon that travels with the panel.
Related reading: our fabrication notes, board quality and design release notes cover the same ground.
Process Control and Verification
On a design of this kind, process capability is the item that decides how the rest of the board is arranged. A stack-up that is drawn rather than described removes most of the ambiguity from a quotation, and it lets the fabricator price the board against the dielectric and copper weights that will actually be used. Where the requirement is not written down, the shop supplies its own default, and the default is chosen for the process rather than for the design.
FAQ
What is a good Cpk? Around one and a third is a common requirement, which corresponds to about four standard deviations between the mean and the nearest limit. Higher values need a very stable process.
Why can Cpk be lower than Cp? Because Cp ignores centring while Cpk includes it. A process that is off centre has a lower Cpk than its spread would suggest.
How many samples are needed? A hundred or more measurements, taken over a period that covers the ordinary variation of the process. A handful of parts gives only an indication.
