Nichia E17A/E21A series (April 19th: updated with output tests in the OP)

Hoop, that makes me think of a board I made for a Luxeon Z once: LuxeonZ, the making of a copper board, and testing
It still works but upon heating up, after a few minutes something deforms and shorts, when cooled down it works again. I think I made the AAA insulating layer too thin.

You could use the DTP pad for probably the cathode, and use enough thermal compound between the star and pill that there is no electrical pathway. It’s not ideal, but better than permanent DD, with a little better thermal characteristics.

It would be nice is someone made a cooper MCPCB with its footprint and a thin or heat conducting insulation layer. The Cree LEDs must have an insulating layer in them, but apparently it conducts heat better than the insulating layers of existing MCPCBs do.

I suppose by PVD you mean physical vapor deposition, that is the relevant meaning in Wikipedia. That and anodization both sound like good ways to make thermally conducting electrical insulation layers. I am thinking thin film diamond, but that is too expensive.

I think a thin sheet of something would be easier to control than a thermal epoxy layer under the star. That would really be the best way, because the area of the insulator conducting the heat is much bigger.

How about the mica insulators used for TO-3 packages on heatsinks in through hole stuff? I dont remember how big a TO-3 is exactly, but i think it would be enough to fit under most stars in this small a light.

I like djozz’s Luxeon Z approach for a diy solution. I would suggest lapping all of the surfaces very flat and then using some thin spacers between the mating parts, like a couple of small pieces of cigarette rolling paper here and there. Clamp the star between some 1-2-3 blocks or other flats for curing; excess epoxy will flow out.

I had bought some MG chemicals high temp epoxy that can tolerate reflow temps but it’s not that cheap. The thermal performance is only ~.265 w/(m·k) tho. Seems timtronics has better performing stuff, around 2.7 w/(m·k).

What about a PCB from OshPark? With the thinner, double-copper boards, and plenty of vias, the heat can be spread across the whole top and bottom of the pcb to help get it away from the emitter. Mask the entire board to make it electrically insulated from the body of the flashlight, leaving only the solder pads for the wires and the emitter. The whole board would be practically DTP. Or, if you’re convinced that only the negative side is carrying heat (although I doubt that can be true, even if it’s designed that way) you could make most of the board connected to the negative side, and only an area large enough to solder the LED+ on that side of the emitter. It should be very simple/easy to design for someone who knows what they’re doing.

Masking the entire board would affect the thermal performance though eh? That’s what these alternative design proposals try to eliminate.

Well, I’m not smart enough to know, but some have said that the junction from emitter to star is the most critical heat path. It does make some sense though, that from the relatively tiny connection point of the emitter, the heat spreads out to an increasingly large area. So, as long as the relatively large outer edge of the board can still conduct heat at least at the same rate overall as the contact area under the emitter, it’s all good, right? Like, if the area under the emitter is 1/10th the area around the rim, then the rim can conduct 1/10th as well and still work fine. I don’t know what the thermal conductivity of solder mask is, so I can’t say if it can do that or not.

Edit: Just for fun, I did some rough calculation. If you consider the entire package size to be the contact area of the emitter, you get 4.41mm² area for the larger Nichia E21A. Supposing a 20mm PCB, and assuming a 1.5mm outside edge contact, you get roughly ~45mm² area. So, like I said, you’d only need the edge of the board to conduct 1/10th as well to be able to carry away the heat as effectively as it is being carried out of the emitter.

I think djozz already came up with the best solution. Battery inserted reversed, battery+ on the entire housing, these Nichias’ anode mounted on the thermal pad (noctigon/sinkpad), connect + and - pads on the board for a cathode; dedicated driver. That beeing said, it’s a dedicated driver what we should consider concerning OSH Park…

Yea that does make sense. I don’t know if it would conduct through the masking 1/10th as well the DTP connection. Comparing high temp epoxies in post #23, there was a 10x difference in thermal conductivity just between two different brands of epoxy. Thermal performance can vary widely among materials.

*Edited for clarity.

Yeah, I just looked up thermal conductivity properties in Engineering Toolbox and the only materials that even come close to 1/10th (or better) are other metals and Diamond.

You could also isolate the MCPCB with a disc of aluminium nitride or something. Aluminium nitride is really interresting for that kind of purposes, but it’s expensive.

Anyone done any testing/measurements with these? I just ordered some 5000K 70cri E21A emitters from futureelectronics and will report any measurements I do.

I ordered some R8000 ones from lumitronics, both 17A and 21A, should be in tomorrow afternoon. I hope to do a test later this week.

Excellent. I’m imagining making my own gapless xhp70 and –50 emitters. Only tests will tell if they can compete in output.

They’re bright and comes in 9080! No tint shift so far.
With XP sinkpad you get oval beam, 2p E21A non isolated, with XM sinkpad you get 4p E21A non isolated. Cathode must be soldered to the MCPCB’s thermal pad. Scrap the masking paint between pads for best result (minimal solder thickness). Also to get the most flat arrays.
Any non conductive thermal pad will do. Best with thin thermal grease on an anodized heatsink.

Yes, 4x 17A should exactly fit an (DTP) XP-pad, all minusses (cathodes) soldered on the thermal pad. The body has become led-minus, a minor problem to solve :party:

And with this one we can get the almost impossible to get: 9080 4500K Nichia
I’ll post my result with 2pcs sm503 and 2 pcs sm303 on a XM DTP board later

It seems there are two ways to deal with the lack of a thermal pad. One way is just to electrically isolate the MCPCB from the flashlight body. As clemence suggested, an anodized shelf with thermal grease should isolate fine, but most shelves are not anodized. Another simple way to electrically isolate the MCPCB is to put a layer of kapton tape around the perimeter of the MCPCB so there is no metal-to-metal contact, then fill the space between shelf and star with thermal paste. Kapton tape is 25-50 microns thick and thermal paste has at least 2 W/mK thermal conductivity, so with a 16mm MCPCB the thermal resistance of the thermal paste is 0.25-0.5 K/W, which is pretty small. Anyway, there are lots of solutions to the problem, but this is the one I was thinking of using.

The other way, as djozz is suggesting, is to electrically connect the flashlight body to the (+) or (-) of the LEDs. I read your description of how to have the LED positive connected to the body, but how would you deal with having the LED negative connected to the body? Putting the driver in the tailcap is one way…