de-doming: illustrations for color changes and increased luminance

Flashlight Modding and DIY Parts
1

Two typical color shifts are known for de-domed LEDs:
- color becomes always warmer (see DrJones explanation, https://budgetlightforum.com/t/-/13495 )
- some LEDs also show an unpleasant green shift

But what system is behind these color changes?

I was so curious that I have inspected a handful of LEDs (six Crees and one Nichia) with a colorimeter (Spyder4 sensor + HCFR software).

No LED color bin

flux bin

(just for completeness)
1 XP-G2 7C4 R3
2 XM-L2 7B4 T3
3 XP-G2 5A2 R4
4 XM-L2 (3A...3D) 0E3 T6
5 XM-L2 (0A...3S) 051 T6
6 XP-G2 1D R5
7 N219B-R85 SW45 D220


The Crees already have been sorted from warm to cool.
I used the nominal current from the data sheets (XP-G2: 350mA, XM-L2: 700mA, N219B: 700mA).

1. color temperature, color "fault"

Here are the results, as grey circles. An arrow shows a change resulting from de-doming.
The labels A, B, C and D65 are basic marks from the software and have no particular meaning here.
The software also prints some basic color temperature values for orientation.

"direction of color shift"
You can imagine a pattern, a swarm of curved lines, which suggests a "direction" of the respective color shift at each point in the diagram.
XP-G2 and XM-L2 seem identical (compare 5 vs 6, and 1 vs 2, I guess it's just the same phosphor.
The Nichia (7.) deviates slightly from the Cree behaviour, but still matches the common pattern.

"length of color shift"
The resulting color also strongly depends on the "length" of the arrow, though.
However, be careful with interpreting the length of the shift, because lines of constant color temperature are askew to
the long, arcuated black body line. And they are more dense towards higher color temperatures at the left.
(For an illustration search for "ANSI C78.377", or see the picture http://en.wikipedia.org/wiki/Chromaticity#mediaviewer/File:PlanckianLocus.png in the wikipedia:chromaticity, http://en.wikipedia.org/wiki/Chromaticity .
The shift for no. 4 seems to be the greatest, but in fact it's just an ordinary shift (see table below).

conclusions?:
Usually de-doming is most interesting for LEDs with the highest flux bins. Inevitably, that's the cool white variants, those with the "thinnest" phosphor. And just these are the most critical. Here you can imagine, which color bins are the most promising for avoiding a unfavourable green shift. The lower the color bin is in the diagram, the better.

For Cree LEDs you can look them up in the PDF XM-Family Binning & Labeling, http://cree.com/~/media/Files/Cree/LED%20Components%20and%20Modules/XLamp/Data%20and%20Binning/XLampXMBL.pdf , especially "ANSI Cool White" on page 7.

You will notice a problem when looking at these vendor data sheets: the area for a color bin usually is quite large. For the same label, a color bin from the "lower end" will be much more promising than one from the "upper end". You won't know at the time of purchase. If you want to be sure and have no better info you might have to buy more LEDs and select the favorable ones. But you will have a good idea which color bins are promising at all.

Here is a table with the numerical values. I haven't bothered to calculate errors, just imagine the grey circles vary at most 1/3rd of their diameter.

No. LED color bin color temp. with dome [°K] ...without dome ..difference CIE x, y with dome ...without dome
1 XP-G2 7C4 2983 2436 547 0.431, 0.392 0.478, 0.406
2 XM-L2 7B4 3061 2469 592 0.421, 0.379 0.469, 0.397
3 XP-G2 5A2 4096 3328 768 0.375, 0.369 0.421, 0.408
4 XM-L2 (3A...3D) 0E3 6745 4037 2708 0.314, 0.298 0.380, 0.380
5 XM-L2 (0A...3S) 051 7869 4959 2910 0.297, 0.305 0.350, 0.396
6 XP-G2 1D 9158 4985 4173 0.294, 0.275 0.347, 0.367
7 N219B-R85 SW45 4470 3678 792 0.360, 0.353 0.391, 0.373


The results for the cool white LEDs appear too cold for me. Also I believe that the LEDs actually are more "greenish" (upwards in the diagram), the colorimeter shows them slightly towards purple (downwards). If so, it's a systematical error, there's no variance: repeated measurements result in almost congruent circles. The values for moderate color temperatures seem to be quite reasonable.

I have a XM-L2, color bin 1C, flux U2 in reserve. It's located exactly between 5. and 6. and I haven't dedomed it yet, in case I find out that I still want to measure something with dome.

2. color reproduction

In contrast to color temperature, you can't tell about color reproduction (CRI) from the above results.
Instead, I made a comparison with a color checker (datacolor).

Conveniently, there are two LEDs which are located very closely:

left picture is no. 3 with dome, center picture is no. 4 without dome. right picture is a mouse-over of both.
(I don't know if that works in all browsers, so I also added the plain images)

I can't notice any real difference. At least for LED no. 4 the color reproduction doesn't seem to suffer.

3. reason for increased luminance?

At first, the following, you might have seen occasionally, seems to be a quite plausible explanation :

De-doming results in a smaller apparent die size (roughly half of the original size).
But as still almost the same amount of light is emitted, "light per area" should be increased now?

But in fact this is not the actual reason! and it conflicts with "conservation of etendue". If that were the explanation, luminance would have to rise also if you removed the magnifying effect of the dome by making it flat (either by squeezing it or by slicing off the upper part). The apparent die size also becomes smaller then, but as a matter of fact the luminance does not increase then. Luminance only increases if the dome is removed.

The following gif animation illustrates this.
It's four pictures of the very same LED, through a sun filter, each with the same current, f-stop, shutter speed and focal length
- with dome
- dome squeezed flat with a thin glass plate (smaller die, but still same luminance)
- upper half of dome cut off with a scalpel (smaller die, but still same luminance)
- de-domed (smaller die, higher luminance)

The pictures were shot with slightly different distances, that's why size varies abit, but luminance is not affected by distance.
Well, with sliced dome ("dome cut even") the luminance seems to be slightly increased. Here, dome thickness is smaller, and I guess that DrJones explanation then applies to a small amount.

[gee, my first article with images, tables and links, what a mess to get all these right, hope I haven't confused other things about this..]

edit 2014-08-04: typo found

2

Fantastic testing!

Worthy of a sticky!

3

I loving this. I plan to dedome some XM-L2 U2 2C. Any ideas what this will look like?

4

Dont. Use emitters located below black body locus (Edit: 1 and 4 A and D tinted emitters). Less chance of getting a messed up green. I have said that for a long time. Still sticking to it and the findings above shows the tint shift towards green too. Even with the tint shift seen from 1As and such, im often not happy with them after de-doming. Often too greenish. But in theory better than de-domed 2C.

Correct tint mixing can fix bad green tint though.

5

I tried dedoming a whole bunch of XML2s. The only one that didn’t look green was the 3A tint. 3A looked by far the best.

6

I was not too pleased with my XM-L2 3A that I de-domed. Only did one though.

There is also some randomness when it comes to de-doming and tints. So If you de-dome 10 emitters, you are more likely to get one or some with a not so bad tint.

7

Great article!

8

Thanks for posting this data and analysis sma. Great work.

9

wow. this is really really good.
Didnt know squeezing dome flat is an option!

Do you think you could test the XML2 U2/1A and XPG2 S2/2B? Since these are the highest output for now?

10

amazing work man

11

I’m looking for an XM-L2 to dedome that would compliment an MT-G2.

12

Amazing test. If this is the quality of your posts, I'm going to have to check out your other 4 posts. I'd love to know how you did the first test. I was looking at some sensors on Sparkfun a couple days ago and thought they might be able to create results like this, which is more incentive to start playing with Arduino.

13

Thanks guys for your positive reactions!

I held the sensor directly towards the LED at a distance of about half a meter, but distance is not critical. I had to pay attention not to twist it at all as the results were drifting then, perhaps due to some effects in the perforated plastic grid front of the very sensor. The sensor gets along fine with the illuminance, even at very high amps. It has problems though with the much higher illuminance if you point at it with a flashlight (that is, LED with reflector or lens), unsurprisingly. More distance, or a good gray card and indirect illumination is required then.

Difficult, especially if the actual chroma coordinate of your LED is at the upper end of "2C".

It might be important that 1A is much cooler and the shift is more high-angle than for 2C.

Best conditions: It's both below the black body curve and relatively warm (that is, low angle and shorter shift).

By the way, while doing the above tests I searched the forums (most results here in BLF) and found positive reactions about all 4C/D, 3A/C, also negative reactions about 3C (surely both a matter of taste and the variation within one bin), as well as negative about all 1* and 2*, and 2B/C, 3B.

I would be happy to test them all if you send them to me :) (should mention that this means germany). I'd send them back to you at my own cost, for reasons of curiosity. However, both color bins don't seem to be "promising"...

14

I have not dedomed all that many xml but found that the warmish purplish ones responded best to my eyes.
good job!

15

Very nice work, thanks a lot for doing these tests! And well represented, thanks for that too :-)

Picturing colour checker cards is difficult I found, you loose a lot of colour information that was there in reality.

16

I've got an XP-E2 R4 1C that I installed in a reflector with a pinhole type opening. The tint is really green. Is it possible that the dome got cracked or damaged at the base to produce such an ugly tint?

The tint is not as green in real life as in the photo, but it's pretty ugly.

I ordered red another XP-E2 R4 1C just to find out if this emitter is damaged or flawed. It's just makes me feel like it's been dedomed with the dome still on.

17

could have gotten overheated too.
had an xml that got run hot enough to solder short shift to the greenish side a bit.didnt seem to hurt it otherwise.

18

Great work sma. Definitely sticky worthy.

So what is the cause of the tint shift? Does the silicone absorb certain colors? Why do the cooler tints shift more? Does anyone know yet?

19

DrJones has made a fine lecture on this subject :-) :

20

Thanks djozz. Read that way back when, but missed or didn't understand this paragraph at the time (from linked thread in djozz's above post):

"Another observation can be explained by this: XM-Ls have quite some angular tint shift, i.e. the light emitted to the side is substantially more yellowish, while with the dedomed LED this effect is greatly reduced. The YAG phosphor has a refractive index of about 1.8. the critical angle agains the dome (~1.5) is about 55°, which means that such a ray to the very side has traveled a 74% longer distance through the phosphor compared to a perpendicular ray and more blue photons get converted to yellow. Without the dome, the critical angle against air is 33°, resulting in only 19% longer distance in the phosphor and thus much less angular tint shift. See these nice images by Tecmo: https://budgetlightforum.com/t/-/11595?page=8#comment-254514 (post #431)"