testing two of KD's new Nichia 219b leds (4500K'92CRI' and 5500K)

Has no one noticed that the R9 of these new emitters is is big fat “0” and nichia doesn’t release R10-R15 numbers? The “9050” version of the new emitters is supposed to have an R9 of 50 but no one as of yet is selling this version. These days I expect to see manufacturers release R1-R15 numbers for their “high CRI” LED’s… Only seems to happen with COB’s aimed at interior lighting though.

Would you explain that in more detail?

The Wikipedia article explains how eight color samples [R1-R8] are employed to calculate the general color rendering index Ra. The article shows the whole range of the color swatches [R1-R15] and talks of better and different standards as well.

CRI and its shortcomings are explained as well on the SHARP website.

Sharp, Xicato, Bridgelux, Soraa, and Yuji are examples of companies talking about and listing R9-R15 scores for their high cri emitters.

As for the new Nichia emitters, the high cri ones show R9 listed as >0, which means better than 0. There is a “9050” version however which is 90+ CRI and with a R9 of 50, which should render red much better. I suppose for us flashlight folk it comes down to tint and the perception of light quality which, you know, what more could you want than a tint that you like and seemingly decent reproduction of colors? I have been getting into indoor lighting for a while though and I guess my tint snobbery has started to cross over. Somebody stop me!

Thanks. Still reading. So far the things that seem to make the best tests of color rendering are flowers. Some flowers are deep red, as I see when they are lighted with a Philips Luxeon deep red led.

Somewhere I saw spectra of various leds, but I failed to find them again later. Clearly flat or very smooth spectra are good for most purposes. I read here somewhere the suggestion that brighter than normal color may be desirable for some purposes. It seems that could be achieved by a spectrum that had peaks near the centers of the human primary colors (presuming that those can be identified in some approximate way). Clearly bright reds should be rendered as bright reds, so your criticism seems founded.
The main feature of white led spectra is the blue spikes. I have been looking for things with colors that miss the blue spikes by being more violet or more greenish, shorter or longer wavelength. It seems that the colors of those things should be noticeably distorted, perhaps looking more like the color of the blue spike, but I haven’t found a good example.
I have not even been able to see the difference between the Nichia and Cree leds, other than the appearance of the white itself.

It may have been links I posted where you saw spectra, if that helps.
a few more or less randomly that might help:



http://www.1023world.net/diy/spectra/ (lots!) (much in Japanese I can’t read.)

I really like this emitter even if it doesn't attain 92 CRI. It has a very nice tint and the CRI seems really good to me. I especially like that it has comparable output to an XPG2 R5 up to 3 amps while having slightly lower Vf. I think that if it is proven that the KD emitter is 92CRI, it will no longer be 5 for $13.xx.

sshhht, keep it quiet!!

I just realized I spoke of 2 different emitters in the above post. The xpg2 comparison was the 5500 temp emitter, not the "92" CRI emitter.

The emitter is definitely promising, for sure, and I will probably use it over XP-G2’s in the future, I just thought the R9 was disappointing on the 92 CRI version is all.

What about using color histograms to compare color rendering of objects under various lighting sources? So, you take a control shot under daylight, or an incandescent light or other true high CRI source. Then, you use some software to come up with a histogram from that photo. Then, you take another shot with the exact same camera settings with the light source being tested, such as a Nichia 219B from KD. Compare histograms to see where the CRI is lacking in your test subject. If you want to compare two LED’s to each other, you do the same thing. This way, camera settings and eye sensitivity or human tint preference wouldn’t matter. It would be pure data. It would not actually tell what the CRI value is or should be. Although, with a chart of known CRI values to compare to, you could “calibrate” your equipment such to get really close. It would be most useful for comparing two sources to each other. It could also be used to evaluate one’s tint preference. Compare the histogram of a photo taken as illuminated by your favorite LED to a histogram of the same shot taken in natural daylight with the same camera settings to see how your preferred tint differs from natural light.

What ruins it to some extend is the primitive rgb-detection of the camera, it does not detect wavelengths directly. A bit more sophisticated way is use a spectrometer, like the simple and cheap, but effective one that bigmac_79 used in his Imalent review. Real CRI can not be derived from those spectra either because the values are not direct photometric power values (but pixel values of the camera whatever they represent), but you can do the trick of dividing two spectra you want to compare, so you get rid of the absolute values and just see the difference relative to each other.

Yeah, that last part, about comparing the spectra, is what I was talking about. If the camera settings are the same, then the limitations of the camera would be a wash when comparing any light source to another with that same camera, even daylight. Basically, it would be like what we’ve talked about here before with luxmeters. They all have limitations, but for comparing light A to light B, the limitations are a wash, because the same limitations show up in readings from both A and B. Calibrating against known good numbers will help a little with real world accuracy, but there’s really no such thing as making it perfect.

I suppose I need to just buy one of the new Nichia’s and subjectively determine if I think colors look real good under it. They probably do.

Just ordered the qty 5 deal on the 4500 B - didn't read thru this test til now, outstanding, so much better than the old ones - nother great job djozz! (Match re-incarnated)

I, respectfully, take exception to "Match Reincarnated". djozz has over 2000 more posts than Match, has sacrificed many emitters to the diode gods, has invested much more time in testing and cost in equipment, has gathered and published much more precise data, and is still here contributing. djozz is not Match reincarnated. He is djozz and stands in no ones' shadow.

Sorry for being overly sensitive about it, but djozz deserves serious props without comparison to others. Being the humble, friendly guy that he is, we will likely disagree with me. That just makes him all that more special.

+1 - Match on steroids? Super Match? Smile

djozz - "stay thirsty my friend"

He can identify UFOs

He can open a Piñata with a wink and a smile

He can speak spanish in Russian

He is left-handed and right-handed.

He is the life of parties he has never attended

He is the only man to have ever aced a Rorschach test

He is the reason those nine ladies are dancing

He lives vicariously through him self

He never says something tastes like chicken, not even chicken

He once challenged his own reflection to a staring contest on the third day he won

He can measure a light's output without turning it on

I got more... Smile

Don't forget, he bowls overhanded.

Darn! Missed that one...

So my Prometheus beta qr copper has a small black triangle with 2 dots around the square the led sits on.
So would this be The 219a right?.
Someone mentioned the QR now has the 219b which has spooked me so just need to find out confirm if the led in my qr is still an A.
I have also written to Jason hui.