Tint measurements

LED Flashlights – General Info
1

(edited: More lights and converted to narrower specified ANSI bins as used for XP series.

I have made tint colour measurements on my flashlights to try verifying the diversity of tints I have observed.

Theese lights are all bought in the period sept. 2010 - Feb. 2011. The tints could very well change when new batches are sold from the vendors.

My instrument is a Colour Analyzer normally used to adjust the white point on a TV set to the right color temperature.

Here is a reference to the Cree ANSI color bins.

What is interesting is the distance from the "neutral line" through bins from 0A thru 5D (See the chart below). Cree also uses the term: 'ANSI Neutral White' for the range 3700K-5300K.

Lights that measure above this line will look a little green/yellow and below the line a little purple/red. In the sceme below, the column "d" show this tint 'offset'. A displacement of 1 can barely be seen, but 10 can easily be seen at least in comparison to another light.

[Technical:] (for the color theory interested the unit of measure here is "CIELUV". The color coordinates xc and yc from the CIE 1931 colorspace is transformed into the CIE 1976 color-equi-distant colorspace as u' and v' and the euclidian distance (90 degr. angle) delta-E to the black body curve (neutral) is calculated and multiplied by 1000. Also look at the explanation to the graph below).

As an example Romisen RC-i3 with d=16 is rather greenish and TrustFire Z1 with d= -6 is purplish although they have almost the same correlated temperature of 5800 and 5700 deg. K.

have fun.

Light symbol xc yc bin temp d

The lights are here placed in the Cree color bins.

[Technical:] The chart is transformed from the CIE 1931 color space (also used by Cree and which can be seen here) into the better 1976 color space called CIEluv (which can be seen here). The advantage with the latter is that the perceived difference between two colors (two points in the diagram) is the same all over the chart for the same distance between points. Each bin is about 6 by 6 CIEluv units (varies some).

2

Thanks for the info! I wish more readings like this were available. One problem is that one copy of a light might have one tint and another could have something different, but better to have limited information than none at all.

3

All my lights are bought since septemper 2010. I have now ordered extra of Ultrafire WF502B R5 and C3 SS in the hope that they have more with that beautifull tint - at least untill I get mine

4

I remember a great deal of effort trying to understand colour theory I discovered that that stuff was very, very hard. In a previous existence I sorted Macs for printers and was always asked about calibrating them properly.

My getout was to tell them I'd want £2000 per machine and that there was no point in calibrating screens anyway - you had to calibrate workflows. So that the image that came into the building was the image that came off the printing presses once the ink had dried.

When viewed in daylight.

Once you'd spent a fortune on daylight bulbs, certified neutral grey paint, calibratable Barco monitors At £9,000 (10600 euros, US$14000), each and they will only stay in calibration for three years. No LCD screen can be colour calibrated properly unless they've seriously improved since 2005), hoods for all the monitors and neutral grey smocks for the operators. The appropriate colourimeters and so on. I still have the documentation for calibrating an Agfa imagesetter somewhere. All 80 kilogrammes of it.

That also involved about 4 days of no productivity when the printers would not be allowed to adjust the presses to get it right as those adjustments would put the whole thing out of calibration. Since printers are skilled people they weren't at all keen to be handcuffed - nearly caused a strike when I told them that.

I've thought of getting colour temperature meter (Standard kit for studio photographers) and building a box out of Kodak neutral grey cards - but the good ones cost a lot. Kodak neutral grey cards aren't cheap either. And I'd need about 20 of them. Duct tape ought to do the rest....

What I'm saying is that I'm impressed!

5

Would it be possible for you to get an LC-Yellow (solarforce drop-in) and measure that? It's a $10 readily available neutral/warm-ish tint that some here already have and would provide a good reference for that part of the spectrum.

http://cgi.ebay.com/Solarforce-Warm-Colour-P4-Cree-led-bulb-4-Surefire-/280534988227?pt=US_Flashlights&hash=item41513035c3

6


For the moment I would'nt bet my money on my measurements in the yellow area. After I have found a solution for calibrating the analyzer, we can take the matter up again.


I don't have a host for the plug-in right now but have a Ultrafire 504B R2 in the HK post. So it could be possible.

At least I have a plan for calibrating. It involves taking measurements of a perfect overcast sky at noon over a few days and averaging that. That should give me the analyzers interpretation of standard white D6500 and from that I can calculate a correction matrix.
It would be very nice to have a well defined emitter for testing.

7

Maybe some 5mm LEDs? - they tend to be more narrowly specified. They also tend to be cheap.

Or some "daylight" CFL bulbs - the ones used for treating SAD are quite closely specified. I'd guess anyone supplying stuff for professional photographers might have useful sources that don't cost too much.

8

Thank's for the tips. I'll try to find some specs. It will have to be white leds as the accuracy of the analyzer is optimized in that area. The purpose for the instrument is to monitor white point adjustments on television sets .

For the moment I'm trying to refine my charts. Excel 97 don't provide data etikettes, so I will have to do some Visual Basic macro that fixes it. How was it now......

9

I have now found a solution to make my measurements more precise and have corrected the results in the start of this thread and added a few more lights.

The results has moved about a bin towards lower temperatures but the tint offsets are about the same as before (the greenish light is still greenish and the purple is still purple (thanks heaven )).

10

How are you correlating the color to black body temperature? Are you just using the cree's conversion or your own?

If using cree's, those emitters look a bit bunched in WD/WG when we know a significant portion are WC's.

I've thought about doing something similar but using RGB->CIE conversion from a camera pic with all post-processing off. The main difficulty I suppose is also calibration, as I suspect just doing white-card balance is not enough.

11

Calculating the correlated temperature from xc,yc is difficult, but I use a ready made calculator from Radiant Imaging which can be downloaded from here

I don't know the Cree conversion. Must have a look.

I know it look strange that bin WC is not the center of the bunch but for the moment I trust the calibration because of the spot-on measurements on the overcast sky at noon, which consistently gives the same value, and the incandescent lamps that lie on the black body curve.

You can't use RGB from a camera as you introduce unknown filter coordinates. If you knew them, you would have to use matrix manipulation to get to CIE coordinates, but you don't know the coefficients in the matrix.

You have to use tightly specified X, Y and Z filters (as in my Analyzer) to have exact measurements. From them xc is X/(X+Y+Z) and yc is Y/(X+Y+Z).

X, Y, Z filter curves can be seen on the SPD Graph in the calculator.

12

Thanks for the link to the SW.

What I meant by "cree's" conversion is that there's a chart similar to yours above in cree's xr-e bin & tint spec sheet with the exact same BB on CIE overlay. But after looking into it, the CCT conversion used is completely standard (Judd/CIE1931) and only based on apparent color (scalar).

On the camera issue, how close do you suppose varying levels of cal get to an accurate value? For example, if I photographed the flashlight in a dark room (daytime ~6500K white balance?), then did RGB->CIE conversion from a relatively unadulterated bayer demosaic, what kind of error are we talking about here on the CIE scale?

For conversion I assume somehow the image rgb is approximately sRGB (or I can get it there) and this works http://en.wikipedia.org/wiki/SRGB#The_forward_transformation_.28CIE_xyY_or_CIE_XYZ_to_sRGB.29

---

Hmm.. sRGB is 5000K, so whatever's 5000K white balance in the camera, close to "flash" setting if there's one, and between fluerecent and daylight.

13

I can see that the Wiki article gives some formulas for transforming from sRGB to XYZ. I have no prediction of the obtainable accuracy, sorry.

To give you an idea of how far two colors are from each other, the 1976 CIE chart is constructed so that the perceived color difference between two points in the chart is about the same everywhere for the same distance, deltaE (calculated in the upper, right corner of Rad. Img. window).The least percievable distance is about 0.001

You can select two points with mouse buttons left-right and read the distance.

I wish you good luck with the experiments!

14

I just got my XP-G R4 from KD with neutral tint bin 4B which I actually bought for using it as a reference for my Color Analyzer. If I could verify that it is really a bin 4B then my tint measurements would gain in trustability.

The test current was about 200 mA with the star lying flat on the table (the black spots have still not left my eyes) and the analyzer above it. It really was a nice, neutral color.

The measurement gave color coordinates x=0.357 and y=0.362 which is on the middle of the border between bin 4B and bin 4A (4A lies beneath 4B), practically on the neutral black body curve. This is in fact better than inside bin 4B (which is on the greenish side of neutral).

The color temperature was 4640K, a very pleasant white.

Of course it could be bad calibration that placed the result on the lover border of bin 4B, but as the majority of my previous measurements was in the greenish area I was expecting a result in the upper part of bin 4B. I conclude that I should not change anything, calibration-wise.

I already ordered more of this nice LED but they seem to be on backorder. Hope that re-stocking don't lower the quality!

In January I will measure XM-L T6 6000K from DX, they should be on backorder too, I fear.

15

Great work you are doing.Good to know about that emitter also, mine is in Chinese's hands who know where, maybe since feed a camel is cheaper than feed a Tupolev they are using again the silk road.

You have another pòssible reference here, a XR-E Q4 5A

http://kaidomain.com/ProductDetails.aspx?ProductId=11123.

Don't worried for the backorder notice, tha's the usual there, two days ago when mine XM-L was ordered I got the same , today is in packing status.

16

Thank's for the kind words, for the link and for the interest in tint. I have unfortunately not had much response in this thread. The tint is often a draw of luck on budget lights, perhaps therefore it has not the big interest, but then there is modding and closely specified LED's as the ones we are talking about.

A perfect light should also have a nice tint.

I was just about to order the XR-E Q4 5A (which should be at 4200K) , but let it be. I do not yet have the tools, nor the courage to make LED transplants. The smallest part must be a star.

A tint measurement only takes a few seconds, then a calculation in Excel, and I have the color coordinates. A third party program then gives me the color temperature and a look in Cree's dokumentation gives me the bin placement.

17

It used to cost so much money to pick a good tint that most people didn't bother as it cost way too much. Look at the HDS guaranteed tint lights. (Now Ra Lights)

Almost indestructible but not all that bright and very, very expensive.

Actually, this thread is very valuable to me- I just don't feel I can add anything to it. This stuff is difficult!

18

"I was just about to order the XR-E Q4 5A (which should be at 4200K) , but let it be. I do not yet have the tools, nor the courage to make LED transplants. The smallest part must be a star."

Don't be discouraged,in fact it's easiest than solder the wires to the star.

I made a post about it and has a very low response also, strange since it's a key issue in the mod's world.I was doing some research ( i do not know if this is the proper word) related to heat against light output, very interesting findings also which i keep for my self till some light meter arrive.My P31 perform better at 1A than 1,4 A due a star thickness issue, the RC-G2 R2 with original star work very fine even gross over drived.There are many differents thicknes stars around there , it's not just an issue of 14,16 and such...

http://budgetlightforum.cz.cc/node/894

Edited, the tips in the original post...

19

Tint measurement of XM-L T6 star tint 1A from KD.
and XM-L T6 6000K star from DX.

I used a black plastic bucket which I matted with sand paper and put the LED at the bottom facing upwards surrounded by a pair of black socks to trap the yellow/green side spill from the LED. This was essential for measuring the center beam tint correctly. Without the socks the yellow side spill would find its way to the sensor and twist the measurements in a yellow/green direction. On top of the bucket I placed the color sensor facing downwards.

The measurements (my power supply can only deliver max 1A). Vf at higher currents include voltage drop in my leads, sorry.
Vf cx cy temp(K)
A: 30 mA 2.57V 0.318 0.333 6200
B: 50 mA 2.60V 0.317 0.332 6250
C: 100 mA 2.68V 0.317 0.331 6250
D: 350 mA (2.90V) 0.315 0.325 6350
E: 700 mA (3.09V) 0.311 0.321 6650
F: 1 A (3.25V) 0.309 0.319 6750
H: 1 A Hot 0.307 0.312 7030

And now for the color Bins, reference is here.

greenish

Bin: 1A 1D
+----------|----BA----+
| | C |
--|- FE | D -|-< Black body
|H | | curve
+----------|----------+

purplish

Orientation: The Black Body Curve going throug the middle of bins 1A and 1D is neutral in the green/purple direction (up/down in the picture). Measurements F,E and D thus have no greenish or purplish content. Towards the left is bluish and towards the right is yellowish.

Conclusion: The tint at 700mA is very close to the center of the promised bin 1A.
At low currents the tint moves into bin 1D and take a greenish direction.
At high currents and temperatures the tint moves towards violet.
The changes are small but easy to see in comparison with another light with almost the same bin.

[Edit] Notice that the measurements presented are for DC current drive. For PWM based dimming the shift in tint was negligible, less than 0.002 on xc and xy moving slightly in violet direction with higher current due to the heat developed during the measurement.

The yellow/greenish side spill could be responsible for the yellow/greenish edge of the center spot reported by many reviewers and perhaps also the rumours about XM-L T6 having greenish tint (of course they could also be of another batch).

On my doorstep is the XM-L T6 6000K star from DX. I will give it the same treatment and report back here.

[Edit new]:

The DX XM-L eventually came, 48 days after purchase. This was expected to be a little warmer (bin 1D) than the $3 cheaper one from KD. But not so, here are the results.

The measurements (got some more juice on my power supply):
Vf(ok) cx cy temp(K)
A: 50 mA 2.60V 0.315 0.323 6400
B: 100 mA 2.63V 0.313 0.322 6550
C: 350 mA 2.73V 0.311 0.319 6670
D: 700 mA 2.81V 0.310 0.317 6780
E: 1 A 2.88V 0.308 0.315 6900
F: 2 A 3.03V 0.303 0.302 7500
H: 2 A Hot 0.297 0.297 810

greenish

Bin: 1A 1D
+---------|---------+
| | |
neutral->-|- - B| A - -|-
Bin: 0R 0U | E DC | |
+-------+-------+---------|---------+
| H | F |
| | | purplish
| | |
+-------+-------+

For $3.37 more I got a colder LED! I must say that the one from KD was the better deal.


20

sixty545 yes please do this also with DX XM-L T6 declared as 1D tint. Your discovering somewhat confirms findings of a SPAMBOT in tint poll thread - http://budgetlightforum.cz.cc/node/1146#comment-18293

So it looks like it's good idea to purchase tint of one class above desired one. Specialy if you want to drive it hard. For example order 2B if you realy want 1D. But of course it depends on batch you get it from.

I'll wait for your findings and make a resume in poll thread.