On page 13 you can see some thermal simulations of various PCB constructions.
The cooling system here is still different from what we see in flashlights but actually much closer than yours:
PCBs are attached to a 4 mm thick alu plate, much larger than PCB
the entire plate surface is cooled but only poorly; heat must travel across the entire plate to be dissipated.
So: there some cooling directly below the LED but most of the heat is dissipated to the sides. Good enough.
Please note that different pictures have different scales, so you can’t compare colours between pics.
Conclusions?
The results are actually quite similar to what you show in that the PCB temperature is quite uniform all across the surface.
Unlike the test above, the PCB temp is very close to that of the underlaying shelf. They used 40 mm² PCB with 5W LED. It’s not very far from VR21SP4 and 2A test.
Does it prove that heat flows across the PCB? Not really. It doesn’t seem to disprove it either though. And I find it interesting in general.
The best cooling method available at “reasonable” cost is those with DTP + thin film ceramic dielectric. But still fairly expensive compared to DTP + epoxy fiberglass IMS.
FYI, in Maukka’s test, the 2A and 2,5A were for 1 LED. That means it was close to 30 watt
Nichia is really investing in their 3030 portfolio…
there’s a new LED again, 757H-F1.
It’s 3V, CRI 70-9050 in 3000K-6500K and CRI 8000 in 2000K-2700K, the most efficient 3V variant so far.
