Production decisions
Before adding machine vision, try a template
The first proposal for checking chip position in cards was a camera on every machine. The customer wanted something simpler. A nest, a lid and a transparent template solved the problem and opened the door to a whole family of fixtures.
When a quality problem appears, the first response is increasingly technological: a camera, a vision system, artificial intelligence.
Sometimes that is the right answer. But it is not always the first one to try.
The problem: where the chip ends up
In a contact card, the chip module sits in a cavity milled into the card body. If that cavity is offset, the module is offset with it.
And that has consequences. The pins in the encoding head, and later in the reader, must rest on the module's conductive areas, which carry power and signals. If the module is outside tolerance, the pins may lose contact or make unstable contact. The result is encoding errors, interrupted communication or cards that do not work in a reader, even when the chip itself is perfectly fine.
That is why it is worth checking the cavity position before inserting the module. At that stage, a deviation can still be caught in time.
How it was checked
The original method was manual: the operator slid a transparent reference sheet over the card, moved it with their fingers until the edges lined up, then compared the milling with the reference.
The problem with that method is not the sheet. It is having to align it by hand every time. If alignment is slightly off, the comparison is no longer valid: you can accept an offset card or reject a good one. And alignment depends on steady hands, time pressure and who is doing it.
The larger proposal and the alternative
The first proposal was to install a machine vision system on every machine to check chip position.
Technically, it made sense. But the customer was looking for something else: something simpler, straightforward to use and with a modest investment.
So the question changed. It was no longer "how do I automate this inspection?", but "what is actually going wrong in the inspection already being done?"
And the answer was clear: the problem was not the operator's eyesight, but the alignment of the reference. If the template always stayed in the right place, the operator could compare at a glance.
The solution: keep the reference in place
The fixture brings together two parts:
- A nest that positions the card. It only fits one way and always sits in the same place.
- A guided lid that holds the transparent template. When lowered, its stops position it exactly where it needs to be relative to the card.
Using it comes down to three steps:
- Place the card in the nest.
- Lower the lid.
- Compare and decide whether it is within tolerance.
The template has the tolerance limits marked on it, so the operator does not measure: they look to see whether the milling is inside or outside. The decision is still theirs, but now it is based on a reference that is always correctly positioned.
That is poka-yoke applied to positioning: the point is not for the operator to align it correctly, but for the fixture to prevent incorrect alignment.
An extra detail: the fit also checks
There is an advantage we were not looking for that turned out to be very useful.
Because the nest is made to fit the format, a card that does not fit, or fits with unusual clearance, is already a warning. It may indicate a dimensional deviation in the card body itself, worth checking before going further.
As well as providing a milling comparison, the fixture checks the format without anyone having to measure it.
What happened next
The first fixture was intended to check milling in one particular section. It started circulating between workstations: other sections requested it for their own checks, and it ended up being used to set up machines and inspect production too.
That demand led to a complete family: individual fixtures, a multiple fixture with interchangeable filters, a backlit one for seeing the inlay through the paper, another for checking continuous web and a set for the inspection station.
And when automation did become necessary later on, it was built on a much clearer foundation: we already knew what to look at, which tolerances applied and why.
When vision is needed
This is not an argument against machine vision. There are cases where it is the right tool:
- When the volume makes manual inspection impractical, even by sampling.
- When one hundred per cent of production must be inspected, rather than a sample.
- When the defect cannot be seen with the naked eye, even with a good reference.
- When every part needs to be recorded automatically.
But if what you need is for someone to check a sample reliably, or set up a machine, the problem is often the reference rather than their eyesight. And that can be solved with much less.
The takeaway
Before automating an inspection:
- Look at what is going wrong in the existing inspection. Often it is the method, not the person.
- Fix the reference in place. A nest that positions the part and a template that does not move remove the main source of error.
- Mark the tolerance on the template, so the operator compares rather than measures.
- Use the recess as a format check.
- And automate when needed, already knowing what you need to look at.
At 7Axis Industrial, we design and manufacture custom inspection and poka-yoke fixtures, from manual templates to automatic systems. We start with what the problem needs.
View the complete project → Quality, inspection and poka-yoke →