AOI Programming, False Calls and Escape Rate

Automatic optical inspection finds what its program has been taught to find, and it reports what its rules have been set to report. The escaped defect and the false call are therefore both consequences of the same decisions, and a program is judged by the balance it strikes rather than by how much it detects.

What the Program Is Built From

A program is built from a library of component images, a set of rules for each joint type, and a board image that the machine compares against. The library decides whether a part is recognised, and the rules decide whether what is seen is accepted or rejected.

Poor library images are the most common root cause of both failure modes at once, because the machine rejects a good part and accepts a bad one when its reference does not represent the process. Our AOI notes describe the machine side of the arrangement.

Teaching From the Process, Not From the Drawing

A library image taken from a good board captures the process that made that board, including its natural variation. Where the library is taken from one golden board, every board that differs in the normal way becomes a false call.

The practical method is to teach from a set of boards that spans the process window, so that the machine learns the range rather than a point. Our placement capability notes describe how that range is measured.

False Calls and What They Cost

A false call costs the time of the operator who reviews it, and a program that produces many of them stops being reviewed properly, which is the real damage. The escape rate rises when attention falls, so a program that is tuned only for sensitivity produces a line that is less safe than its detection figures suggest.

The fix is usually in the library rather than in the threshold. Raising a threshold suppresses the false call and suppresses the defect it resembles at the same time, so the two have to be separated by a better reference image.

Escape Rate and How It Is Measured

An escape is a defect the machine accepted and a later station found. Measuring the escape rate requires the later station, which is usually X-ray or a functional test, and it requires the results to be correlated with the same board rather than counted separately.

Correlating the two is what makes the figure meaningful, and it is also what identifies the rule that failed. Without the correlation, an escape is known to have happened but the rule that allowed it is not. Our joint criteria notes set out the acceptance the rules are written against.

Lighting and the Limits of Optics

Optical inspection sees the surface. A joint that is hidden under a package, a solder ball that is joined at its base, and a fillet that is shadowed by a tall neighbour are all outside what the camera can prove.

Recognising that limit is part of building a program, because a rule that guesses at a hidden feature produces calls that cannot be resolved. Where the feature is hidden, the inspection route is X-ray and the optical program should not be asked to cover it.

Programming Time and Maintenance

A new product needs a program, and the time that takes is a real cost that is often left out of a quotation. Reusing a library between similar products is what keeps that cost down, and it is the reason a shop keeps its library organised rather than per product.

Maintenance matters as much as the first build. A program that is not reviewed after a process change will start producing calls that reflect the old process, and the operator will learn to accept them.

Changeover and Program Selection

Changeover is a program load plus a verification that the correct program is running. The failure mode is a board inspected with the previous program, which produces either a flood of calls or a silent pass.

Verifying the program against the board rather than against a label is a cheap control. A first board checked by an operator against the drawing catches the wrong program before the line runs an hour with it.

Records and Feedback

The records worth keeping are the call rate per product, the escape rate correlated with the machine that found the defect, and the library revision. Those three describe whether the program is getting better or worse.

Where a defect escapes, the investigation is faster when the image is retained, because the rule can be re-run against the picture rather than against a description. Our board quality notes describe how the standard is applied across stations.

Process Control and Verification

Running a first article through the same checks as the production panel confirms that the two agree, and that comparison is the cheapest form of process control available at prototype stage. Keeping a sample from the panel turns a dispute into a measurement, because the same coupon can be re-examined by both parties without rebuilding the batch.

Reviewing the design before the data is released is cheaper than correcting it after the panel is in the tank, because every step downstream inherits the decision made at the front end.

Checks Before Release

The checks that matter are the ones performed on the product rather than on a sample kept for the purpose, because a coupon that travels with the panel is the only evidence about that panel. Where a requirement can be measured, it should be measured at the point of manufacture and recorded against the board or the lot it applies to.

FAQ

Can AOI replace visual inspection entirely? It replaces the routine part and it cannot judge a feature that is hidden or that needs a decision the rules do not contain.

Is a lower call rate always better? No. A call rate that falls because the thresholds were opened is a loss rather than an improvement, and the escape rate is the figure that shows which happened.

What does gopcb provide for AOI? We provide program development from the customer data, libraries built from boards that span the process window, correlation of escapes with the station that found them, review after every process change, and records of call rate and library revision per product.