Light Bulb CRI_Grades

btw, it’s intentional that a R9080 scores a C+. Compared to what’s possible in the future, CRI=90 and R9=80 is kind of lame.

You make it sound like it's an objective comparison even though we can't be sure of that.

Tint is often more important than bare CRI

XPL HI V2-3D is a lot better than some CRI90 green ugly tinted Chinese COB

Also efficiency plays a big role in a rating
there are some CRI80 COB LEDs that have awesome tint and 8500lm from 50W

BLF is so nerdy. :nerd_face: Ok. I’ll discuss my CRI_Grade calculation when I get back to the computer around noon.

Would you use a CRI98 light that’s ugly green tinted just because it gets A+?

I’m back.
Ok. So CRI_Grade. The convention of marking “R9080” to mean Ra=90 and R9=80 works pretty good as a judge of CRI quality. With today’s tech, R9 is the hardest to achieve, and that puts it in plain view. However, some manufacturers now know how to trade R12 for R9. This means R12 could be the only one that is super low and go unnoticed. Now, I’m also a believer in simplifying things, so this is what I did…

I marked my readings, for example, R908060 (Ra=90, R9=80, R12=60). Then I just averaged those numbers. In this example, the answer is 76.66. This gives great importance to the weakest areas of LED technology, resulting in a super useful number. But grading still needs to be done, otherwise it’s not simple enough. So I just applied my High School’s grading system. This specifically calls for rounding to the nearest whole number, then categorizing. (Grade table in spreadsheet). So in this example, 77 = D+.

To see the formulas, grading scale, and to play with your own numbers, download my spreadsheet here

Now if you can read a data dump from a spectrometer, this obviously has less information. But as a muggle, this is like having that data dump read for you. And it focuses on the hardest parts of making an LED, so if the manufacturer gets this right, it’s doubtful they could AND would screw anything else up too bad.

It seems on BLF, silence is approval :smiley:

I was kinda hoping for some reactions after taking the time to compile that post…

Except for Duv and angular tint shift, both of which are extremely important and noticeable.

Edit to add: I don’t think I’ve ever seen bad tint shift in a household bulb because they’re generally quite diffuse, but it’s common in flashlights. Look at the beam from a flip-chip Cree like XP-G3 or XP-L2 in a reflector and see all the rainbow colors.

Duv is harder for companies to keep under control. Because our vision is so sensitive to green light, adding just a bit of green can bump the lumens way up without wrecking CRI, in order to recapture some of the efficiency penalty you normally find with high-CRI LEDs. But doing this also gives the light a green tone (a positive Duv), which deeply annoys a lot of people.

I don’t want to say CRI_Grade is anything more than a color rendering score… But at these strict of levels, I don’t know of any real life bulbs that would have bad Duv or tint shift while scoring good on CRI_Grade.

Whats the advantage of this over something like CQS or TM-30?

About CRI rendering: recently, I started to appreciate the newest CRI measuring method, which uses two values to present overall CRI. Rf (fidelity) is very similar in meaning to Ra (I won’t try to research for the differences now), and the other value is Rg (gamut), which tells, that how much the examined light source is greenish/yellowish (values under 100) or rosy (values above 100).

Rf measures as the same as Ra (range 0-100).
Rg measures from (100 minus Rg’) to (100 plus Rg’), where Rg’ = 100 - Rf

I would use a starting formula something like this:
0.7 × (Rf / 100) × Rf + 0.7 × ((100 - Rf) / 100) × Rg’’ + 0.15 × R9 + 0.15 × R12

Rg’’ is calculated like this: 100 - (abs(100 - Rg) * (100 / Rg’)) - this tells us, how much the light source deviates from the black body line, if compared with other light sources having the same Rf.

So, for example, if an emitter with Rf=90, has an Rg of 93 (could be called very dull or greenish) or 107 (could be called very ill or rosy), has an Rg’’ score of only 30, while light source spot on BBL would have Rg’’ = 100.

The first two Rf based coefficients corrects for the cases, where e.g. Rf = 98 and Rg = 102 -> that is definitely not a bad light source, which would deserve Rg’’ of 0.

Edit: Please ignore above anything I said related to Rg, as of maukka’s clarification.

Speed and simplicity would be the advantage. Can you imagine if I posted all 41 of the TM-30 (or CQS) diagrams and your job was to spot the best ones? Or sort them?

That route crossed my mind, but where are you going to get all that data? It’s hard enough finding CRI data alone. Plus, Rg and Rf aren’t the same sensitivity as CRI, like you noticed, so fixing that is iffy and untrustworthy imho.

Yes, I know, that this is hard to get, but all light sources, which was tested by maukka, have these data.

Like I say, it’s no replacement for reading a data dump if you can (and have it). But it sure is slick as an alternative grading method.

TM-30 Rg describes how saturated colors seem on average. It’s not directly correlated with duv (magenta/green tint).

Here’s a link to TM30-15 Excel calculator which you can use if you have SPD data from a spectrophotometer.

Oh, good to know. Maybe more good examples would be needed here for my better understanding.

Still, basically, anytime I see an Rg above 100 in a measurement, the duv reading is negative and for Rg under 100, the duv reading is positive. Would be good to know, how indirect the correlation can be.

The inherit limitation as you’ve noted is the CRI Ra being limited the 8 pastel colors being tested. With something like TM-30 including a wider and deeper sample base the Rf means a lot more than CRI Ra.