We tested the fully assembled device and unfortunately the optical power dropped to 40% after a short duration.

This was unacceptable and we needed to investigate. Probing the voltage across the current sense resistor showed the driver was working correctly.

The only possible reason seemed to be a thermal problem. When ordering the LED panel as copper base PCB, one needs to define thermal heat sinks. These are areas of the PCB that are full copper and can provide excellent thermal dissipation. Unfortunately in this version, we had not defined the heat sink pad of the LED as direct heat sink. To test this hypothesis we modified one of the test PCBs by milling out the heat sink pad and mounting a fitting copper heat sink into this slot.

And indeed, the optical power now was stable.

So we ordered a new version of the LED panel. The white areas in the layout are direct heat sinks.

We then desoldered all 100 LEDs and resoldered them to the new panel.

And finally the test in the fully reassembled device showed much more stable optical power output.