A true-wireless earbud is dozens of delicate steps around parts the size of a grain of rice: magnets that must face the right way, adhesive that must cover but never spread, solder that must not reverse a battery, and a left and right that leave the line as a matched, tested pair. Most of those faults disappear the moment the shell is pressed shut — so the automation has to inspect at every step, not only at the end.
A true-wireless earbud is one of the harder consumer products to build well. It packs a battery, a speaker, a microphone, an antenna, magnets and a small circuit board into a shell the size of a fingertip, across dozens of placement, bonding, soldering and test steps — and it is built twice, as a mirrored left and right that have to end up a matched pair.
The line began largely manual: semi-automatic dispensing, hand fixtures, and inspection by magnifier and eye, mirrored for both sides. That scales badly and it judges inconsistently, but the deeper problem is that many of the faults it is looking for cannot be seen at the end. A reversed magnet, a battery soldered the wrong way, adhesive that has crept onto a microphone or a contact — once the shell is pressed and the glue cured, none of it is visible.
So the inspection cannot wait for final test. It has to sit at the steps that create the hidden faults, and each unit has to carry its identity, its pairing and its electrical, acoustic and cosmetic results all the way to the box.
The ones that decide whether the line can be automated without shipping hidden faults, stated before any of them had an answer.
Magnet direction, battery polarity, adhesive overflow onto a mic or a contact — once the shell is pressed and the glue cured, none of it is visible. Inspection has to happen at the step that creates the fault.
Many steps dispense and cure adhesive. Too little fails adhesion; too much reaches a contact, a microphone or an acoustic path. Judging coverage without touching the part is a vision problem, repeated at every dispense.
Magnets, diffusers, flex antennas, speakers and batteries are small and easily damaged. Handling and placing them repeatably, without misalignment or harm, is as hard as inspecting them.
Left and right ship as one paired unit, and a fault can be electrical, acoustic or cosmetic. Identity, pairing and test results all have to follow each unit to the box.
The bonding, vision, test and pairing blocks are marked because the line’s whole logic is to catch a fault at the step that creates it — before the shell seals it in — and to keep each unit and its pair traceable to the box.
The scope of the work, from mapping the manual process to an automated and inspected line, rather than a capability list.
Rarely the subsystem that sounds difficult. Written out because a reader facing the same line gets more from this than from a list of what went well.
The highest-value inspections are the ones at the step that creates a hidden fault, not at the end. Placing vision and sensors at each of those steps, at line rate, was the core of the work — and it is what a final-test-only line cannot do.
Verifying that adhesive covers what it should and nothing it should not, from images, across many different dispense steps each with its own rule.
Grippers, fixtures and motion tuned so tiny, fragile components are placed repeatably, without damage or the small misalignments that fail later tests.
Carrying each unit’s identity, pairing and test history from sub-assembly to the packed box, on a line that runs left and right in parallel.
A unit passes only if cosmetic, mechanical, electrical and acoustic checks all pass. Combining those into one clean pass-or-reject per unit, without ambiguity, is its own piece of work.
The faults that seal inside the shell are caught at the step that creates them, rather than inferred — or missed — at the end.
Magnifier-and-eye inspection replaced or backed by machine vision that does not tire or drift across a shift.
Electrical, acoustic and cosmetic results tied to each unit and its matched pair, all the way to pack.
A line built to be reconfigured for the next product rather than rebuilt, because consumer audio changes often.
Customer projects are presented at property, capability, outcome and integration level. The customer, the product and the commercial terms are not named. Internal architecture, process documents and confidential deliverables are not disclosed, and where a detail would identify a customer it is omitted rather than approximated. More is available under a non-disclosure agreement, within the limits the customer has agreed.
Every item links to its own page, with characteristics, applications and maturity stated honestly for that item.
Feeding, dispensing, test and sort.
SOLUTIONThe per-step inspection stations.
PLATFORMThe machine vision platform.
PLATFORMControllers, test capture and traceability.
CASE STUDYInspection and sorting at volume.
KNOWLEDGEHow the handshake and record work.
Because the most costly faults — a reversed magnet, a battery wired the wrong way, glue overflowing onto a microphone or a contact — are invisible once the shell is sealed. The only place to catch them is the step that creates them, so inspection is distributed across the line rather than concentrated at final test.
With machine vision. The dispensed adhesive is imaged and its coverage judged against what should and should not be covered. It is a per-step check, because the line dispenses adhesive at many stages and each has its own pass criteria — enough to hold, never enough to reach a contact or an acoustic path.
Each unit carries an identity that is scanned and tracked from sub-assembly onward, with pairing and test results bound to it. The left and right builds run in parallel, and the pair is confirmed matched and jointly tested before it is packed.
Yes; it is built to be reconfigured. Fixtures and programs are set up to re-tool for the next model rather than requiring a new line, because consumer-audio products change frequently and a line that cannot follow them does not pay back.
Written for engineers. Share it with one.
Tell us the specification, the constraint and the deadline. Programmes that cross silicon, radio, embedded and AI are where Faststream is strongest.