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

Nice testing clemence. The gaps must of been so small that its not untill you take a macro shot that they appear large. :+1:

What I had not hoped for with the 17A and 21A: the actual die under the phosfor is clearly smaller than the full dimension of the led, thus the fading edges. That will not be solved by filling the gaps with silicone. A quad led set-up will face the same challenges that the first gen. XHP leds (and the old MC-E and MK-R) have: no smooth reflector for this quad!

Silicone will reduce phosphor surface temperature by reducing the photons bounced back and accumulates heat on the phosphor surface. Each time the photons hit air gap, they will bounced back unless in straight line. I didn’t mention it to cure the cross mark.

FYI, they survived 8A (for a minute or two). Still editing the pictures

- Clemence




Well that looks a bit more promising. The picture really shows which parts need better heat sinking. Do you know if noctigon XP16/20 MCPCBs use 2oz copper thickness? It seems like the area/layout of the copper on the MCPCB could make a big difference, too. On noctigon stars the trace from the pos and neg contact is the width of the footprint (3.5mm). On the other hand, on this 20mm TPAD I have in front of me the trace is thinner and there is a bottle neck of 1mm right at the footprint; not as good for heat transfer. Ideally the trace would spread out and cover as much area as possible to increase the heat transfer across the insulating layer to the metal below, just like the Nichia 119 MCPCB that djozz designed.

Based on my thoughts above, I think I came up with a good MCPCB design for good heat transfer from the non-DTP LED contact. Start with a copper disk or sand off all the insulating stuff on a copper MCPCB. Then put down some Kapton tape covering all the area except for the DTP area where the LED cathode will connect. Cover all the Kapton tape area with copper tape; this will be the other LED contact. I have some copper tape that is 2-3 thousandths of an inch thick, which is about 2oz thickness. I made a mock-up of this design on a TPAD MCPCB; see picture.

The copper tape is electrically isolated from the bulk copper below. The Kapton tape doesn’t have very high thermal conductivity, but is very thin, about 1-2 thousandths, which is about 3-4x thinner than the dielectric layer on MCPCBs, by the look of it. When the Kapton tape is spread out over a 20mm MCPCB the extra thermal resistance is only about (40microns)/[(300mm^2)(0.12W/mK)]=1.1 K/W. This should make for good thermal transfer away from both LED contacts. I tried soldering on the copper tape, and the copper tape adhesive got a little soft when everything was at soldering temperature, but once it cooled down it stuck well again.

The Sinkpad has a way better heat transfer from the outer solder pads than Noctigon

if you look right the positive and negative side go all the way over the star like a bar between the 2 wire pads
it looks pretty much like your self made star cut to half width

maybe someone makes a board with insulated base and 3 DTP pads insulated by a dielectric

That would work well, but I doubt someone makes it.

I estimate that, in my design above, the bottleneck is the thinness of the copper tape. Heat will have trouble traveling laterally. Using ~1mm copper shim instead of the tape should greatly improve the lateral heat transfer.

It’s done already. 144A and E21A are exactly what the new VirEnce board all about. Only one problem left: $$$$
Until then, Djozz board is your best option. It has very large copper pour for its size and thick copper base. One problem for extreme BLF application is the dielectric used is only 3 W/MK.
A single E21A can be driven really hard > 3A on Djozz board. His last test showed that 144A with 4 times the heat performed OK.

- Clemence

I’m planning on a quad configuration, though.

You can do that too….4 pcs in a single row. Just scrape the masking paint. But, no real use in most FL

Great posts, Clemence! Very nice job you did there with handling these tiny tiny LEDs.

Wait later until Djozz come with his E17A. I myself is rather not to deal with them

:+1: :beer: I like that tests…

Waiting for Djozz’s test results ……drum rolls……

Yes, me too :frowning:

I have have received the leds and reflowed them (both 5000K R8000) on my boards (on which they are going to be tested until your boards will be available), but I need 2 hours for the actual tests, and 3 more hours for the write-up. I’m not sure where to find those hours in the next week.

Very nice! Take your time, we’ll wait for your test result later.

- Clemence

April 19th: I updated the OP with two output tests on my 119 boards.

Thanks for the test Djozz,
Mine was dying at 8A, but still lit. The damage pattern truly depends on how the pads are cooled. The burnt always started from the hottest point. This LED should create a better nice beam pattern with it’s rounded rectangular hotspots.

Thanks for the tests. I cancelled the order I had placed after I saw that the dies didn’t extend to the edge, and I’m glad I did. These might be fine for some applications, but it’s not what I was looking for.