Poka Yoke in PCB Assembly: 7 Devices That Stop Mistakes
A poka yoke is a device or a rule that makes a mistake impossible to commit, or impossible to pass on unnoticed. The term comes from Japanese manufacturing practice, and on a PCB assembly line the idea is simple: rather than training people not to load the wrong reel, make it physically impossible to fit the wrong reel to that feeder position.
The distinction that matters is between prevention and detection. Prevention stops the error at the moment it is made. Detection catches it later, which is useful but always second best, because a defect that has already been placed has to be removed, and every removal is itself a risk to the board.
Prevention Before Detection
A physical poka yoke is normally a pin, a key, a shape or a stop that allows only the correct action. A reel that fits only one feeder slot, a tray that can only be loaded the right way round, a connector that only mates in one orientation: these are devices that remove the possibility of error rather than warning about it.
Where a physical device cannot be fitted, the next best thing is an interlock, in which the machine refuses to start unless a condition is met. The weakest option is a warning that an operator can dismiss, and it is worth being honest about which of the three a given station actually uses.
Setup Proofing on the Line
Setup is where most wrong part events begin, because the machine is being changed deliberately and the opportunity for a mix-up is highest. Feeder carts that are built and verified offline, with the reels locked to the cart, remove the risk of a reel being placed one slot out during a pressured changeover.
Where the machine supports it, the feeder position can be verified against the program by reading the reel identifier, and the machine can refuse to run until every position matches. That converts a human check into a machine check, which is the whole purpose of mistake proofing.
Scan Verification of a Part Number
A scan verification step asks the operator to scan the reel label and the position, and compares the result with the program. It is a detection device rather than a prevention device, so its value depends on the process stopping when the result does not match. A scan routine that logs a mismatch and carries on is not a poka yoke, it is a record of the error.
Scanning also has to be designed so that it is hard to defeat. Where a single scan clears several positions, or where a label can be moved from one reel to another, the check loses its meaning. Reel identifiers generated by the storage system, rather than typed by hand, are far more robust.
Orientation and Polarity
Orientation errors are the classic component placement fault, and they are best attacked with physical features. Asymmetric package outlines, chamfered corners, offset locating pins and tray cavities shaped to accept a part in one position only all make the error difficult to commit. Where a package is genuinely symmetric, a marking convention plus a defined orientation check in the program is the practical fallback.
For connectors, the housing itself is the poka yoke: a keyed shell will not mate in the wrong orientation, and a cable assembly built with the wrong plug will not fit the header. That protection is lost if the connector is supplied without keying, which is a design decision worth making early.
Board and Fixture Keying
A board that can be loaded into a machine or a fixture in the wrong orientation will eventually be loaded that way. Tooling pins that engage in one direction only, an asymmetric panel outline, and a fixture pocket that accepts the panel in a single position all remove the possibility.
Where an assembly is processed twice, once on side one and once on side two, the fixtures for the two operations should be mechanically different, or the boards should follow a defined route that makes the second pass impossible to confuse with the first.
Process Interlocks and Recipe Control
Machine recipes are another place where a mistake produces a whole batch of defects. An interlock that prevents a program from starting without the correct profile loaded, a barcode that selects the recipe automatically, and a rule that only authorised staff can change a profile all reduce the chance of a wrong setup running for a full shift.
Where a machine is shared between products, the changeover should end with a first article check that is defined by the recipe rather than by habit, and the machine should not be released to production until that check is recorded.
Count and Continuity Checks
Counting is a simple and effective device. A feeder that expects a known number of components for a panel, a tray that holds exactly the number required, or a kit issued complete and returned empty all provide evidence that nothing was left over, and a missing part shows up as an empty cavity that the count cannot explain.
Continuity and presence checks at test then act as the backstop. The scan based checks described in the wrong part procedure and the presence checks in the missing part routine are the detection layer that catches whatever the prevention layer did not stop.
When an Alarm Is Not Enough
An alarm that can be silenced without action is not a control. Where the process depends on people responding to a warning, the warning has to stop the line, or the response has to be recorded against the board, or the control will decay within weeks.
The same principle applies to a feeder that can be left mis-loaded: the feeder setup discipline that keeps reels seated and locked is a poka yoke in spirit, and the placement checks that follow it are the safety net. Prevention at the station and detection at the joint together produce a process that fails rarely and visibly.
Where Human Factors Still Decide
Not every error can be engineered out. Judgement, reading a drawing and interpreting an unfamiliar instruction are all human tasks, and mistake proofing works alongside them rather than replacing them. What the devices do is remove the slips that happen when attention is elsewhere, which is precisely when a trained operator is most likely to make a simple error.
The sensible order of work is to list the errors that have actually occurred, rank them by the cost of the batch they could destroy, and then attack the top of the list with the cheapest device that removes the possibility. A reel that will only fit one slot costs nothing to specify at the machine, and it removes an entire class of event for the life of the product.
FAQ
Is a scanner a poka yoke? A scanner is a detection device rather than a prevention device. It becomes a poka yoke when the process cannot continue until the scan matches, because at that point the mistake cannot be passed on.
Where should mistake proofing be applied first? At the setup of the line, because that is where the largest number of possible errors exist and where a single mistake affects a whole batch. Placement orientation and part number come next.
Can poka yoke replace training? No. A device prevents one specific error, and operators still need to understand what the device protects and what to do when it blocks them. Mistake proofing reduces the consequence of a slip, it does not remove the need for competence.
