An update on the 3x3 array. In summary I successfully reflowed the LEDs, but had burning problems that I suspect was from leftover flux.
After some snags with the thermal adhesive I used to fill the gaps between the fingers I was able to flow the LEDs on the board without much trouble. I put together a hotplate that consists of a piece of 1/4” aluminum plate with a candle (or two) under it and a thermocouple to monitor the temperature of the plate. I tinned the fingers of the copper board. Then, at room temperature, I placed each LED with some flux on its contacts on the board. Then I heated it up until the solder melted. I tried to minimize the amount of flux but there was still quite a bit. I then gave it an agitated bath in warm isopropyl alcohol to try to get rid of most of the flux.
I did some preliminary output measurements up to 9A. Unfortunately the output is just about 9x the output of a single XD16 at the corresponding current, so no significant boost from being in an array. Also unfortunately my reflow was just not as clean as I would like regarding flux. I monitored the LED closely during testing and some edges started to darken and smoke a bit. I suspect it was leftover flux because there was definitely a lot of it that I couldn’t clean from between the dies. Next I will try flowing some LEDs without flux.
Great job testing them EasyB. I hope you can get around the problem soon. Try to use boiling IPA, warm isn’t enough to thin and clean that flux residue.
Possibly a stupid question, but for someone that has the time and know how to quickly produce the eagle files, would it make sense to try an Oshpark fr4 board? Just for optical testing and low cost? Also possibly in 3p3s to minimize current imbalance due to Vf differences? Obviously currents will be limited, but it may actually help determine how heat sensitive these new CSPs are. Just a thought…
An update: I redid the LED reflow without flux and got some good results (5600 lumens at 20A), but observed some slight charring of some of the silicone edges on the LED surface.
Thanks for the tip.
I decided to not use any flux because I think it might be difficult to remove all of the residue and I wanted to give it the best chance of succeeding. I tinned the pads on the board then after it cooled I scrubbed it with paper towel and alcohol to remove the flux. Then I started with 9 fresh XD16s. I placed them on the hotplate next to the board and heated it up. When the solder melted I skimmed the oxide off the solder pads and quickly placed the LEDs on the pad, 3 at a time.
I did some testing and saw the output was up to 10% more than 9x the single LED output at the corresponding current, so it seems like there is some efficiency boost from being in array form. In my first try pictured above you can see some of the LEDs are dimmer which I attributed to Vf differences. With my second 3x3 array attempt all LEDs appear the same brightness. All LEDs are from the same strip so that is a bit strange that the first 9 had a large variation and the second 9 doesn’t.
I went up to 20A which is the limit of my power supply. After that test I noticed some slight charring of some of the silicone edges. I guess this is the “photon crosstalk” so it looks like Clemence was right about this.
My understanding of what’s going on is that the silicone edges are being heated too much as a result of light absorption. I couldn’t get a good picture of the charring, but it is right at the top edge of the silicone, where there is the least thermal mass and thinnest thermal path to the rest of the silicone. This leads to too high a temperature. I think I will try to fill the gaps with clear silicone, like used to seal silnylon tarp seams. I figure the silicone filling will reduce optical boundaries which might reduce absorption, and it would increase the cooling ability near the edges.
Clemence, you had explained some optical benefits of filling the gaps with silicone that I didn’t quite understand. Do you have any recommendations?
My gapless designed achieved roughly 70% of the gapped output with extensive thorough cleaning. Reflow solder - boiling IPA cleaning - reheat (150C) - boiling IPA cleaning - reheat (150C) - boiling IPA cleaning. Too risky to launch a product requiring this procedure. Then I could get shorter process using aggressive water based flux with similar result but damaged the solder mask. Not everyone have access to water based flux.
Bridging the gaps will reduce the heat caused by photon cross talks. I have no reliable data to present here, but I tested it and I can only say it does increase the burning threshold. I didn’t measure the nett total output unfortunately because the max current achieved was too low for me to think it was worth the extra works.
As for the optical silicone filler…You can get them from either Shin-Etsu or Dow Corning. But to make it work you still have to pass the flux problem in the first place. Then the silicone also needs gap (0,05mm) to start filling properly under atmospheric condition because it’s rather thick. If you want to close the gap further (and also to prevent micro bubbles), you’ll need a vacuum chamber to make sure all those poured silicone filled the gaps. DIY vacuum chamber is easy to make but again, not everyone would care to do so. Too much hassles for small gain.
I think DIY gapless LED matrix is not for everyone but it’s DOABLE.
I sliced out the small sections of silicone that were charred then covered the surface and filled the gaps with silicone. I used permatex “flowable” windshield repair silicone. After letting it cure for a day I again tested the LED at 20A. After a couple minutes I saw another spot in a gap that started to burn. I again carved the dark areas out and tested again. The LED was fine for 5min at 20A so I decided to go ahead and put it in the emisar D1S.
Above is a beamshot of the D1S with 3x3 XD16 (left) compared to a EE X6 with dedomed XPL V6 1A. These two lights have similar max beam intensity, and in this picture the center beam intensities are the same for both lights. The D1S has roughly twice the beam diameter. You can see that the D1S doesn’t have a donut hole, but it does have a dark ring around the bright beam center which results from the 4 crosses that surround the center die. The beam from a reflector takes an angular average of the emitter around its center, and since there is a larger proportion of dark area at the radius where the 4 crosses lie, a dark ring is produced.
With a VTC5A at 3.97V, I measured 99 Kcd. With the same partially charged cell a ceiling bounce test showed that the D1S was about 3.4x brighter than the X6 with dedomed XPL V6 1A at around 6A, so roughly 3.4x1650lumens=5610 lumens. This measurement was taken at probably around 5seconds after turn on. The output drops quite quickly after turn on.
So, I’m pretty happy with the result. The 3x3 XD16 array allows for a very powerful single-cell light-weight flashlight. A dedomed XHP70.2 would have similar output and luminance, but it requires two cells. A triple or quad dedomed XPL or SST40 would also give similar output and effective luminance (depending on how efficiently the reflectors fill the head of the light), but this usually requires a making a custom spacer and results in a heavier flashlight. But the real unique advantage of the XD16 array is its scalability. A 6x6 array would have ~4x the output of the XHP70.2. I’m making tentative plans to make a 6x6 array and put it in an astrolux MF-02. :laughing:
Nice work! It’d be great to see them perform up to binning specs and there to be a 3x3 array mcpcb as well for these. Competitive peak intensities, tons of lumens, and huge hot spots is exactly what got me excited about this LED.
Great work, thanks for checking out the options for the XD16.
It is amazing that you got the project as far as a very well performing flashlight, with all that fiddling with tiny leds and a self-made 9x9 array ledboard
But I’m not sure if you will have many followers considering that the performance is not miles better than other leds, and making a 9x9 array, even if a dedicated board is available, is more work with more room for error than reflowing a single led.
Thanks JaredM and djozz. There is definitely a lot of room for error. Some more spots started to burn after some use in the flashlight. The burn spots are always right at the top edge of the silicone where two LEDs meet. When I carve out the burned spots the area seems to be stable and not burn anymore. If I redo this array with fresh LEDs I might preemptively slice these top corners off.
I also discovered that the die does not cover the entire footprint. I measured 1.60mm across for the package and 1.38mm for the die. I updated the OP with this information and a picture.
The new F2 by Evitek is supposedly made with the XD16, They get a custom 1x3 version of the chip.
1.6 mm is the diameter of the halogen filament. Three in a row is the length of the halogen filament.
see my thread over here…. look at picture at post #9.
I see that the 1000 lumen output is based on calculations for lambertian emitter. And that EasyB later found out that the die is actually smaller than silicon, so the surface luminosity is likely not uniform. Scaling the output down for 1.9044mm² emitter gices me ~750lm. The true output is probably somewhere in between…
The lumen calculation is not dependent on the surface area, just the candela above the LED and assuming the angular distribution is Lambertian. Sort of like how the spill from a light is proportional to the total output of the LEDs… See the post I linked in the first post.
You’re right, thank you for correction.
Then this LED is not so bad…much higher output than E21A despite much smaller size. Though I can’t help by wonder how much of this improvement can be attributed to better cooling afforded by soldering the LED on such custom board.
Even if tint is greener than that of Nichia, arrays can fix that.
knew it!
