We had used the same 365nm LED source since about 2018. Reliable enough, but the wavelength binning drifted, some batches came in centred more like 368 or 370nm and for forensic customers that matters. So in the middle of 2024 we decided to move to a different supplier who promised tighter binning, 365 plus or minus 2.5nm guaranteed.
On paper it was a clear win, tighter spec and a slightly lower price per thousand. What could go wrong. Well.
The new LED had a different thermal pad layout. Our existing aluminium heads were machined for the old footprint, so the heat path was not as good. First production batch ran fine on the bench for ten minutes, then under continuous use the output dropped because the junction was getting too hot. Customers running long inspection sessions would have noticed the beam fading.
We caught it because one of our assembly guys, Old Zhou, mentioned the heads felt hotter than usual after the burn-in test. He has been doing this longer than I have and his hands are a better instrument than half our meters. We pulled the batch before anything shipped, which was the right call but it blew the delivery date for a German customer by nine days.
Daisy had to make that call to Germany. They were not happy but they understood once she explained, and honestly they would have been a lot less happy receiving torches that dimmed after fifteen minutes.
The fix was re-machining the head to match the new footprint and adding a bit more copper under the LED for heat spreading. Took us about three weeks and a re-tooling cost we hadn't budgeted. So the cheaper LED ended up costing more for a while.
Now the wavelength is genuinely tighter and the customers who care about that are getting better torches, so it was the right move in the end. But the lesson I keep relearning is that you can't change one part in isolation. An LED is not just an LED, it's an LED plus the metal around it plus the driver. Change one, re-test all of it.
