Last month a Tier-1 automotive customer in Suzhou pulled a batch of dome sheet arrays off their SMT line and found two panels out of every 500 had a single tilted snap dome - tilted roughly 12 degrees off center. The domes still actuated, but the shifted peak put the button dome outside the rubber plunger on their steering-wheel module, producing intermittent double-clicks on cruise control inputs. Root cause traced back to a lamination fixture that had drifted 0.15 mm over three shifts, and our human operators, inspecting at 120 panels per hour under a magnifier, had missed it because the defect was below the 0.1 mm visual resolution threshold at that throughput. We have spent the last six weeks fixing that gap.
Today, Shanyo is commissioning a new automated AOI (Automated Optical Inspection) station on the dome sheet array lamination line in our Qingdao plant. The station runs at 36 panels per minute, takes eight-megapixel images per panel under angled ring lighting, and catches six defect classes that human inspection historically could not hold at throughput. Outgoing quality on the affected line has moved from 99.5% to 99.8% first-pass yield over the first two weeks of pilot runs, with false rejects held below 0.3%.
AOI station scanning a die-cut dome sheet array panel before reel packing.
What the AOI station actually inspects
The machine is not a generic PCB AOI repurposed for domes. We specified it around the failure modes that actually show up on metal dome production lines, based on 18 months of 8D reports from our customers. Six defect classes are covered:
- Missing dome - a pocket with no dome seated. This happens when pick-and-place suction drops mid-transfer; historically a 0.02% escape rate, now zero on the inspected line.
- Tilted / rotated dome - angular offset from the reference center, measured in degrees. The Suzhou steering-wheel issue was a 12° tilt; the machine flags anything above 3°.
- Adhesive misalignment - the PSA layer shifted relative to the pocket pattern by more than 0.2 mm. This is the root cause of most "peel-and-stick fell off" complaints in consumer remotes.
- Dust / fiber contamination under the dome - a 50-micron fiber visible as a dark speckle between dome and PCB pad. This is the leading cause of intermittent contact in die-cut dome modules used in medical panels.
- Scratched or cracked dome - radial cracks in the stainless steel snap dome. Catches cracked SUS301 domes that would fail within 5,000 cycles instead of rated 1 million.
- Wrong dome variant shipped - vision reads the laser-etched lot code and cross-checks against the PO. We once shipped a 250g force batch when the customer ordered 180g; this stops that class of human error.
Why we stopped trusting "eyeball at 120 panels an hour"
Manual inspection of dome arrays has not changed much since the 1990s: an operator sits under a ring light, magnifier at 5x, flipping panels to check each pocket. The physics of human vision cap useful detection around 0.1 mm at a 30-second inspection cycle, and attention degrades after about 40 minutes on the same task. Our quality records showed that 78% of customer-reported defects in 2025 fell below that threshold - they were not operator errors, they were cases where the defect simply was not visible to the human eye at line speed.
The AOI camera uses a 20 μm/pixel resolution across a 350 mm field of view, which means it can resolve a 40 μm fiber or a 2° tilt in a dome that a human would need to stop and measure under a microscope. Ring lighting at 45° angled, plus dark-field backlight, gives contrast on specular stainless domes that flat illumination cannot.
Pilot numbers, first two weeks
We ran the station in "shadow mode" for two weeks - inspecting every panel but not yet pulling rejects - to validate that the algorithms matched what our QA lab would classify. Numbers from 4,200 panels inspected (about 1.8 million individual domes):
| Metric | Before AOI | After AOI (pilot) |
|---|---|---|
| First-pass yield | 99.5% | 99.82% |
| Customer-returned defect rate (ppm) | 1,800 | 320 (projected) |
| Inspection speed | 120 panels/hr | 2,160 panels/hr |
| Minimum detectable defect | ~100 μm | ~40 μm |
| Operator labor per shift | 2 inspectors | 1 (tending the machine) |
The station does not replace final outgoing inspection - we still pull a statistical sample per IATF 16949 and run it through a force-curve tester - but it has moved the bulk of repetitive, resolution-bound checking off human eyes.
Four-legged snap domes after AOI pass, before reel packing.
What this means for our customers
If you are sourcing custom dome sheet arrays for automotive, medical or industrial control applications, the practical change is three things:
- Tighter incoming-quality agreements. We can now offer a 500 ppm outgoing-defect ceiling on automotive programs, with full traceability back to the specific AOI frame that inspected each reel. This makes PPAP submissions much faster - the AOI log is an automatic measurement system.
- Lower assembly fallout. Customers running SMT lines tell us that tilted domes and missing domes are the two defects that stop their line entirely, because they require a human to open the reel, rework the pocket, and re-seat the panel. Cutting those by ~10x shows up directly in their OEE numbers.
- Faster sample-to-mass ramp. New programs historically took two to three weeks to dial in lamination alignment because feedback loops were slow. The AOI produces a defect map per panel - tilt angle distribution, adhesive offset distribution - which lets our process engineers tune the fixture within hours instead of days. This pairs with our existing FPC and dome array rapid-sample work.
What the AOI does not do
Worth being clear about, because we see customers over-automate based on marketing: the AOI catches visual defects. It does not measure actuation force, click ratio, or lifecycle. Those still go through the force-curve tester and the 1-million-cycle endurance rig in our lab. We see AOI as a complement, not a replacement, for the mechanical testing that actually defines consumer-electronics dome performance.
It also will not catch a wrong-specification material - if SUS301 is swapped for a cheaper stainless at the incoming-material stage, the AOI cannot tell. That is why we still do material-certification checks on every incoming coil.
Next steps
The station is live on our main dome array line now. We are commissioning a second identical unit for the silicone-over-rubber button line by end of Q4, and a third smaller unit for prototyping samples so that even our 48-hour rush samples get the same inspection rigor as mass production.
If you are currently qualifying a Shanyo dome array program and want to see the AOI station running - we can do a 15-minute live video walkthrough during your factory audit, or share a sample AOI report from your last shipment. Contact our engineering team to request a defect report sample or to book a virtual line tour. We will send a real reel inspection log, not a marketing PDF.
- John Smith, Senior Application Engineer, Shanyo
Last month a Tier-1 automotive customer in Suzhou pulled a batch of dome sheet arrays off their SMT line and found two panels out of every 500 had a single tilted snap dome - tilted roughly 12 degrees off center. The domes still actuated, but the shifted peak put the button dome outside the rubber plunger on their steering-wheel module, producing intermittent double-clicks on cruise control inputs. Root cause traced back to a lamination fixture that had drifted 0.15 mm over three shifts, and our human operators, inspecting at 120 panels per hour under a magnifier, had missed it because the defect was below the 0.1 mm visual resolution threshold at that throughput. We have spent the last six weeks fixing that gap.
Today, Shanyo is commissioning a new automated AOI (Automated Optical Inspection) station on the dome sheet array lamination line in our Qingdao plant. The station runs at 36 panels per minute, takes eight-megapixel images per panel under angled ring lighting, and catches six defect classes that human inspection historically could not hold at throughput. Outgoing quality on the affected line has moved from 99.5% to 99.8% first-pass yield over the first two weeks of pilot runs, with false rejects held below 0.3%.
AOI station scanning a die-cut dome sheet array panel before reel packing.
What the AOI station actually inspects
The machine is not a generic PCB AOI repurposed for domes. We specified it around the failure modes that actually show up on metal dome production lines, based on 18 months of 8D reports from our customers. Six defect classes are covered:
- Missing dome - a pocket with no dome seated. This happens when pick-and-place suction drops mid-transfer; historically a 0.02% escape rate, now zero on the inspected line.
- Tilted / rotated dome - angular offset from the reference center, measured in degrees. The Suzhou steering-wheel issue was a 12° tilt; the machine flags anything above 3°.
- Adhesive misalignment - the PSA layer shifted relative to the pocket pattern by more than 0.2 mm. This is the root cause of most "peel-and-stick fell off" complaints in consumer remotes.
- Dust / fiber contamination under the dome - a 50-micron fiber visible as a dark speckle between dome and PCB pad. This is the leading cause of intermittent contact in die-cut dome modules used in medical panels.
- Scratched or cracked dome - radial cracks in the stainless steel snap dome. Catches cracked SUS301 domes that would fail within 5,000 cycles instead of rated 1 million.
- Wrong dome variant shipped - vision reads the laser-etched lot code and cross-checks against the PO. We once shipped a 250g force batch when the customer ordered 180g; this stops that class of human error.
Why we stopped trusting "eyeball at 120 panels an hour"
Manual inspection of dome arrays has not changed much since the 1990s: an operator sits under a ring light, magnifier at 5x, flipping panels to check each pocket. The physics of human vision cap useful detection around 0.1 mm at a 30-second inspection cycle, and attention degrades after about 40 minutes on the same task. Our quality records showed that 78% of customer-reported defects in 2025 fell below that threshold - they were not operator errors, they were cases where the defect simply was not visible to the human eye at line speed.
The AOI camera uses a 20 μm/pixel resolution across a 350 mm field of view, which means it can resolve a 40 μm fiber or a 2° tilt in a dome that a human would need to stop and measure under a microscope. Ring lighting at 45° angled, plus dark-field backlight, gives contrast on specular stainless domes that flat illumination cannot.
Pilot numbers, first two weeks
We ran the station in "shadow mode" for two weeks - inspecting every panel but not yet pulling rejects - to validate that the algorithms matched what our QA lab would classify. Numbers from 4,200 panels inspected (about 1.8 million individual domes):
| Metric | Before AOI | After AOI (pilot) |
|---|---|---|
| First-pass yield | 99.5% | 99.82% |
| Customer-returned defect rate (ppm) | 1,800 | 320 (projected) |
| Inspection speed | 120 panels/hr | 2,160 panels/hr |
| Minimum detectable defect | ~100 μm | ~40 μm |
| Operator labor per shift | 2 inspectors | 1 (tending the machine) |
The station does not replace final outgoing inspection - we still pull a statistical sample per IATF 16949 and run it through a force-curve tester - but it has moved the bulk of repetitive, resolution-bound checking off human eyes.
Four-legged snap domes after AOI pass, before reel packing.
What this means for our customers
If you are sourcing custom dome sheet arrays for automotive, medical or industrial control applications, the practical change is three things:
- Tighter incoming-quality agreements. We can now offer a 500 ppm outgoing-defect ceiling on automotive programs, with full traceability back to the specific AOI frame that inspected each reel. This makes PPAP submissions much faster - the AOI log is an automatic measurement system.
- Lower assembly fallout. Customers running SMT lines tell us that tilted domes and missing domes are the two defects that stop their line entirely, because they require a human to open the reel, rework the pocket, and re-seat the panel. Cutting those by ~10x shows up directly in their OEE numbers.
- Faster sample-to-mass ramp. New programs historically took two to three weeks to dial in lamination alignment because feedback loops were slow. The AOI produces a defect map per panel - tilt angle distribution, adhesive offset distribution - which lets our process engineers tune the fixture within hours instead of days. This pairs with our existing FPC and dome array rapid-sample work.
What the AOI does not do
Worth being clear about, because we see customers over-automate based on marketing: the AOI catches visual defects. It does not measure actuation force, click ratio, or lifecycle. Those still go through the force-curve tester and the 1-million-cycle endurance rig in our lab. We see AOI as a complement, not a replacement, for the mechanical testing that actually defines consumer-electronics dome performance.
It also will not catch a wrong-specification material - if SUS301 is swapped for a cheaper stainless at the incoming-material stage, the AOI cannot tell. That is why we still do material-certification checks on every incoming coil.
Next steps
The station is live on our main dome array line now. We are commissioning a second identical unit for the silicone-over-rubber button line by end of Q4, and a third smaller unit for prototyping samples so that even our 48-hour rush samples get the same inspection rigor as mass production.
If you are currently qualifying a Shanyo dome array program and want to see the AOI station running - we can do a 15-minute live video walkthrough during your factory audit, or share a sample AOI report from your last shipment. Contact our engineering team to request a defect report sample or to book a virtual line tour. We will send a real reel inspection log, not a marketing PDF.
- John Smith, Senior Application Engineer, Shanyo
