Light Bulb CRI_Grades

In all honesty, I think that is not very correct. If required for an average, one or two typical mean R12 values could be obtained out of a bunch of Maukka's tests from typical standard high CRI emitters like Samsung LH351B/C/D, Nichia 219B/C, Luminus SST20, maybe some Osram Oslon (like here)… should fall between 70 and 80 points, namely.

I’m not sure what you are saying.
But if I didn’t say it clear before, setting R12 = R9 is a hack you should not trust.

Joshk I mean to say that it would be better to estimate some (one or maybe two) average high CRI emitters R12 value, which should fall around 75 points (?). It would anyway be a different hack, but a bit better because only really good leds have very high R12 scores.

It’s not a bad thought, but anytime you go just typing a number you believe pulls the score in the right direction, the internet WILL accuse you of biasing the data. No matter how good your intentions.

Thanks for this work Joshk! I'll also try to add your bulbs tests results in it today.

Dealing with missing info is a tricky subject. So far on the spreadsheet I choose the strategy of not altering the results (just highlighting the unfilled cells), if the info is missing it is missing. Therefore those bulbs with missing info are left with a disadvantage and can't be "fully compared".

Toward our goal to order bulb light quality I don't see a clear better scientific strategy... Except being sure to have all the data, of course!

In both cases, the Qfactor/CRI_grade are biased, one time for missing data, one time for assuming a supposed value.

Yea I agree, posting estimates is just a bad idea. I won’t be doing that anymore.

But that comes at the loss of now not knowing which unknowns are the best candidates to buy and test. So as a compromise, I may keep using the R12=R9 hack during my sorting step, but then deleting that and the score before posting. That will at least give a hint, without actually giving a score.

As I said before, the R12 = R9 hack isn't really good because R9 scores still vary greatly among high CRI leds and in no way are linked to R12, whereas R12 is fairly predictable. As far as I know any decent led scores above 70 points in R12, even low CRI ones. See what I said above.

For led emitters or bulbs without R12, use an asterisk and a footnote to clarify the sorting dilemma. You could also divide the table listing between emitters/bulbs with R9 and R12 on top, and the rest below (so no need for estimates or hacks).

Yea. I’ll just call them r9080 or whatever when they are missing the R12.

I'm working on integrating your results. Two questions:

  • is it normal in your spreadsheet to have negative values for R9?
  • do you have the formal "brand / model" name of the tested lights?

Yes, if the R9 is negative, that’s still the number I use. I’ll add some more brand/model info when I get time.

For record, most 70 CRI LEDs have R9 in the –20s.

My iPhone flash has an R9 of –59. But it still manages a positive CRI_Grade of 13. I’ve never seen a negative CRI_Grade, though it’s not prohibited.

Wait what?

That has to be a mistake or the lens must be stained so much it distorted the waveform. If the CRI really is that low, it means Apple must be actively trying to source the sh*ttiest possible LED on the market. That really goes backwards, especially for use as camera flash.

Oh well, I didn't read the table, I though you meant Ra of 13. :p

Newer iPhones are actually quite nice


I thought my iPhone 8 was a newer iPhone :weary: :frowning:
Thanks for letting me know. I guess. :stuck_out_tongue:

Interesting…only a generation difference. Unsure of what changed and when

I tested a few Soraa bulbs before and although the CRI and R values were above 90, the DUV was about 0.00 or slightly higher. For me, that made everything too yellow in comparison to my other lights with negative DUV. Even light bulbs that were a good bit warmer temp but with negative DUV rendered whites whiter than the Soraa.

Today I received some Auxma LED light strips I bought after sorting SKV89’s data and they are awesome :slight_smile:

AUXMA 2700K CRI95 2835 DC 12V 120LED/m 9.6w/m
My measurements:
CRI_Grade: A- (95%)
Duv: –0.0017 (–1.2 MacAdams)
CCT: 2626K
Ra: 96.7
R9: 90.8
R12: 98.3
(Also sorted into spreadsheet in OP)

I’m proposing tint scoring as one of the criterion.

Tints inside desirable range will get 100 points. For each .0011 duv away from desirable duv range, the light loses 25 points.

For CCT up to 5000k, the desirable duv range should be about –0.0055 to +0.0000
For 5000k and above, the desirable duv range should be about –0.0022 to +0.0033

You can talk about these numbers. Those are not set in the stone.

CRI_Grade has been massively helpful for me to sort light sources by their most elusive metric. But it’s not a tell-all. I don’t think that’s possible with everyone having different tint preferences.

For easy reading on my part, I did create a separate metric for tint called “MacAdams” (as seen in the spreadsheet) that I didn’t discuss online.
We already know 1 MacAdams ellipse defines the smallest noticeable tint change, so I just set that equal to one. So in the 4000K area of the CIE chart, that is a Duv of 0.0014. So for a light with a Duv of –0.0014, the MacAdams factor is –1.0.
For my preferences, any light between +1.0 and –1.0 will impress me. But some people will can’t stand positive, so they will want to look for 0.0 to –2.0??
It’s almost unnecessary, except I tend to forget how much Duv = 1 MacAdams ellipse. So I find it useful.