GB 351D closed.

I think I have some of those, it mainly requires doing it and that what I do and don’t do in my hobby is quite random. And unlike in my work, no-one corrects me :person_facepalming:

Please add 4 more for me, I already asked for 6 but I’d like to buy 4 more if possible

What I did do:

(perhaps this is not the right thread for this, but I already started it here, sorry about that)

I sliced the 4 LH351D 4000K 90CRI leds in my SP36 sample, and took some measurement before and after.

The sliced leds in reflected light show that I sliced very close to the phosfor, on many area’s all silicone is gone on top of the phosfor layer (and perhaps a tiny bit of phosfor too but that does not show up when looked at the lighted leds)

I should also have done tint tests before the mod, but I forgot them :person_facepalming: . So the tint results before slicing will be represented by the hotspot of a Q8 that I have that has the same leds, that should be similar.

Hotspot of Q8 with intact leds, low 7135 under 350mA current (so under 90mA per led):

Hotspot of Q8 with intact leds, highest FET current (over 4A per led):

Hotspot of SP36 with sliced leds, low 7135 under 350mA current:

Hotspot of SP36 with sliced leds, highest FET setting:

Here’s the beam of the Q8 with intact leds compared to the SP36 with sliced leds next to each other on a white wall. Phone camera white balance set to sunlight, the tints are not completely accurate to reality, but you get an idea. The tint difference is larger in reality than in this picture.

In my own experience the tint went from ok but not really nice (unsliced) to super-nice, rosy and not yellow, nice both at low and high current. The tint improvement is larger than the tint/CRI data suggest.

I did notice some slight rings in the beam after slicing the leds but that is a SP36 thing, has nothing to do with the leds. It may even be just present in the prototype SP36 that I have.

Now for the output loss, I did measure that before and after. As I noticed before in other mods with it, after slicing the 4000K 90CRI LH351D, I lost a lot of output, more than with an oldschool Cree dedome.

Measurements with fully charged 30Q’s.

Before slicing: 3seconds 4250lm, 30 seconds 3940lm
After slicing : 3seconds 3490lm, 30 seconds 3235lm

So the loss is 18% compared to before slicing. One must suffer for getting the perfect tint :slight_smile:

The final cost of changing the SP36 from 70CRI cool white with tint shift in the beam (the stock XP-L2 leds, about as bright as this host can go) to 93CRI 3500K with no tint shift in the beam (the sliced LH351D 4000K 90CRI leds) is 34% light loss. For me that is an ok price payed.

Thanks djozz, quite enlightening shaved dome tint and CRI study for high CRI emitters. It is quite pleasing to see it actually enhances CRI a little bit :-) further. Now I understand why my sliced SK98 XHP50A 4000K CRI90+ looks noticeably warmer than a 4000K emitter in reflector. I like this. I also have another sliced XHP50A 3500K CRI90+ G4 flux bin waiting for action, it's gonna look warm and cozy. :))

Cheers :-)

34% light loss? Since our eyes and brains do not perceive the different visible light spectrum frequencies with equal sensitive one could also say those low CRI emitters are “gamed” as more energy is invested in the frequencies we are more sensitive to, something which can be seen as a sort of cheat which inflates the lumen figures.

I can pleasingly deal with such “loss”. :-D

Cheers :-)

What the loss of output for Cree dedome?
I always assume the loss should be like xpl hd -xpl hi differences. ~3 bin or ~20%.

The few times that I measured it (old type cool 70 CRI Cree XP-E2, XP-G2 and XM-L2), the light loss was close to 10% for a clean dedome. But it may vary, I only measured it a few times.

I remember some measurements from fonarevka, there was about 20% loss for xml2

I can not find back my numbers on the XP-G2 and XM-L2 (I never managed a good dedome of the XP-L) but here is what I found for the XP-E2, it was even a bit less than 10% for these two leds:

Interesting,thanks. It is effectt of geometric optics, maybe it depends of die dimensions.

Thanks djozz!
Where these Samsungs all from AEDs first sale?

The Samsungs from the Q8 were from AEDe’s first sale, but the sliced ones from the SP36 were recently bought from led4power. I should have mentioned that, but I did not see an obvious difference in tint between those two sources (I know because when I bought the l4p ones I hoped that he had sourced a batch with a tint closer to the BBL, but it was clearly not so but instead very similar)

I did not test the slicing of that exact led (I could not find my stash) but I did try it on the 5000K 90CRI LH351C, the SPHWHTL3D50GE4RTMF to be exact, that maukka did extensive testing on (link). The effect of slicing this led should be similar to slicing the 5000K 90CRI LH351D though.

I now did my homework and did do the tint measurements before the slicing too.
I modded a Utorch S2+ clone with the SPHWHTL3D50GE4RTMF, and with a 2.8 A biscotti driver. All tint measurement were done in the hotspot (which is the most relevant part of the beam that you want data of).

First with intact led at 2.75A

Then I sliced the led, not the cleanest slice but I’m beginning to suspect that how close you are to the die has more impact on the led behaviour than how even your surface is.

Now with sliced led at 2.75A

Same sliced led at 20mA (not from biscotti driver but by connecting the flashlight head to a led tester, so not PWM-ed)

Now apart from the colour data, that show a very similar effect as the slicing of the 4000K version (lower colour temperature, preservation of CRI, closer to the BBL), the beam looks like a very good neutral tint: no green, not too yellow, not too rosy, I like to call good coloured neutral tints “brownish”. Here is a beam shot for an impression, all lights are at fairly low level, the sliced LH351C beam is in between a (led=intact) LH351D 4000K 90 CRI beam (Q8), and a 219B 4500K R9080 (ROT66, beam from TIR optic).

(the Q8 is a bit too green, the ROT66 is a bit too red, the picture shows less tint difference than reality)

I also measured the light loss from the slicing of the led and I now got –19%, so that is also similar to slicing the other Samsung led.

One thing not mentioned before: slicing causes better tint and less light, but of course it also decreases the spot size of your beam and increases throw (did not measure by how much, my educated guess is 35–40). This can be relevant to what flashlight you are building.

Thanks a lot for these efforts djozz.

If you feel like it, you could also do some staged slicing measurements in some other emitter. I mean, to see how does it compares a not so close slicing versus a close one with regards to output, hotspot size, temperature and CRI. Should be a gradual regression/progression in those parameters, I believe.

Cheers :-)

Yes, that will be interesting too. So many options for testing…
Tomorrow my work week starts again, little chance for testing then.

Thanks for the tests djozz! Looks like there will be some Samsung shaving in my future.

djozz, could you give a link to technology of such dedoming ? Want to try it too.

Thanks for the testing Djozz. Now we know :+1: :wink: .
I’ll definitely try a dedome in the next mod or two I do. 4300k is right up my alley.

Thanks djozz, this is very helpful information and makes me happy that I put this LED into my headlamp.