English
| Client | Tier-2 optical module manufacturer (name withheld under NDA) |
|---|---|
| Location | East China |
| Production | 100G / 400G transceiver TOSA and ROSA subassemblies |
| Equipment | CG-CT5-20050-1200W water cooled LED UV curing system, 200 x 50 mm window, 365 nm |
| Result | Cure cycle 45 s to 12 s, first pass yield 92% to 99.2%, temperature rise below 10°C |
The customer builds TOSA and ROSA subassemblies for 100G and 400G data center transceivers. Each module passes several UV bonding steps: the collimating lens is fixed after active alignment, the fiber array is potted onto the planar lightwave circuit, and the photodiode die is attached to the submount. Adhesive layers are thin, but they must reach full strength before the module moves on, so the cure step sets the pace of the entire line. Their line used a 6 kW mercury lamp over an indexing conveyor, and yield was being lost exactly where the module should have been finished.
- Slow cure became the bottleneck. Each module needed 45 seconds under the lamp, capping the line at 60 modules per hour.
- Lamp heat moved the optics. The mercury lamp radiates infrared, so the submount and lens barrel warmed during exposure and settled after cooling. About 5% of modules drifted out of alignment tolerance and had to be reworked or scrapped.
- Unstable intensity. Output decayed over lamp life, forcing recipe re-validation every few weeks and a fresh process qualification after each lamp change at roughly 1,000 hours.
- Cleanroom load. Ozone generation, a 6 kW heat load and 1.8 m of occupied line length all worked against the production team.
We supplied one CG-CT5-20050-1200W water cooled LED UV curing head with its matched chiller, mounted directly at the output of the alignment station instead of a separate cure tunnel. The head delivers 18 W/cm² over a 200 x 50 mm window at 365 nm, matching both the photoinitiator system of the customer's optical adhesive and the pitch of the module tray, so a full row cures in one pass.
Output is dimmable from 0% to 100% through a constant-current supply, and the head is interlocked with the conveyor PLC, so exposure starts only when the tray is in position. Working distance was set at 30 mm and verified with a radiometer at nine points across the window before the first production run.
- Adhesive and wavelength audit. We reviewed the adhesive datasheets with the process engineer and confirmed 365 nm as the deepest through-cure option for the lens and fiber array adhesive. 385 nm and 395 nm remain available for future inks and coatings.
- Line trial. Twenty modules were cured at 100% output for 12 seconds and all twenty passed the customer's shear test for adhesion, cure depth and alignment drift.
- Recipe lock and ramp up. Two recipes were fixed: 12 seconds at 70% for lens bonding, and 20 seconds at 50% for fiber array potting, where slower cure reduces shrinkage stress. Two weeks of 100% inspection was followed by AQL sampling.
| Metric | Before (6 kW mercury) | After (1200W LED CT5) |
|---|---|---|
| Cure cycle per module | 45 s | 12 s |
| First pass yield | 92% | 99.2% |
| Substrate temperature rise | about 45°C | below 10°C |
| Alignment drift out of tolerance | 5% | 0.4% |
| Curing station power draw | 6 kW | 1.2 kW average |
| Lamp service life | 1,000 h | 20,000 h |
| Warm-up time | 5 minutes | instant on / off |
| Curing station footprint | 1.8 m | 0.6 m |
Output rose from 60 to 180 modules per hour at the same headcount, and the alignment rework loop nearly disappeared. Because the head draws power only while the tray is in position, metered energy use at the curing station fell by about 80%.
- Through-cure in one pass. 18 W/cm² at 365 nm cures the full adhesive layer in 12 seconds, so the head sits inline at the aligner rather than feeding a separate oven.
- Cold light protects alignment. No infrared output means the lens barrel and submount stay dimensionally stable during exposure, the main reason alignment drift dropped.
- Repeatable intensity. The constant-current supply holds the set level across a full shift, so a recipe validated months earlier still matches production.
- Window matched to the tray. 200 x 50 mm covers a complete row of modules, and the window can be customized for other tray layouts.
"We were worried a 1200W LED head would not cure our adhesive as deeply as the mercury lamp. The radiometer data and shear test settled that in one afternoon. What we did not expect was how much alignment yield would improve once the heat was gone." — Process Engineering Manager
If your line cures optical adhesive or fiber array potting, send us your adhesive type, substrate and target cycle time for a trial recommendation.