Color Rendition Index (CRI) - Tutorial and questions


I’m trying to understand CRI on LED.
I figured it could be of interest to people here so I will explain a bit what I learnt, then I have some questions for Pixelblaze applications.

I understand CRI has its limitations and is a bit obsolete (TM30 seems to be the new norm) though, but I’ll keep on it as I found no TM30 measurement tool within my budget.

Color Rendition Index (CRI) is the capability of a light to restitute colors accurately, as “accurate” means “under sunlight”. A 100 CRI light is “perfect” is basically sunlight, which restitute all colors well, including red and skin colors.
Due to the fact that artifical lights do not emit photos through nuclear fusion, all the light wavelenghts are not emitted as balanced as the sun, resulting in different color rendition.

For example on this picture we can see that cool white LED barely emit red color, which will result in red objects, and skin tones to appear blue/green.
In the case of adressable RGB LED, the main colors emmitted are blue, green and red, but red LED is usually weaker in terms of energy, so it results in a quite poor white light, that feels very cold, not very homey.

In order to have a nice white LED light, it is recommended to use white LED (which are actually blue LED with red phosphor) for a better quality of light, though there are differences, of course between the LED.

In my case, i’m using SK6812 RGBW adressable LED strips. They contain a red LED, as well as a white phosphor LED in the same package. I bought a small Opple Light Master 4 that measures CRI up to R14, which basically means it calculates:

  • R1 to R8 (average is Ra) → this value lacks red and skin tones so it’s not really a good indicator of white light quality
  • R9 (red rendition) → this usually lacks on RGB lights
  • R10 - R14 > additionnal tones

Now let’s compare 3 measurements I did:

  • LED Rope (WS2812 RGB led inside):

    For some reason, this is very good including in red so I’m a bit puzzled by this sensor result. If anyone know why ? But we can see the addtionnal R>10 are not so good.

  • SK6812 RGBW in full white:

    Pixelblaze, if requested full white, will light up only the white LED of each module (RGB are OFF).
    As you can see, the light output is good (473 lux), but it is pretty mediocre on the R9 (Red), which makes it actually a pretty poor choice for quality lighting, despite having a white LED.

So my idea was to compensate with the RED LED from RGB package: this will obviously warm up my white, but this is not an issue for home lighting.

  • SK6812 RGBW in RED desaturated at about 50%

    Here I get a very good result (and a nice warmer light) accross the range, which shows that using white LED and red LED works nice to improve “on the measurement” the quality of the light.
    But the problem is that Pixelblaze does not allow discrete control of the white and red LED, which means if my RED LED is at 50%, the white LED is lowered at 50%: I can’t have both LED at 100% power at the same time, resulting in lower light output (262 lux).

Now I have several questions on this if any one can help me:

  1. Am I actually making sense (for someone who has more experience than me on this) ? Is it realistic to say that I can compensate the lack of red color of my white LED by adding a red LED on top ? Are measurements using this “cheap” tool are accurate enough to deduct this ? I’m a bit cautious because the cheap RGB LED rope I bought seems to have pretty high quality and it is visibly NOT as good as the RGBW SK6812.
  2. Is there any way I can control white and RGB LED separately with Pixelblaze ? It seems WLED can do it (and fried my fuse at the same time, it gets HOT !). Because turning off RGB when the white LED seems actually to offer lower light quality than white + red.
  3. Is there a way to improve light quality other than pushing next to my RGB channel a full white channel with dedicated high CRI white LED ? And if I have to do this, is it controllable directly from 1 output of the Pixelblaze V3 ?

Hope this wasn’t too long !
Thank you !

I wouldn’t fully trust that device or results. We know the spectral spread of RGB is terrible, there are 3 distinct spikes.

White LEDs (such as used in RGBW addressables) are much better, using a phosphor to emit a board spectral range, but still has spikes that won’t match sunlight.

Consider this graph from an SK6812 RGBW LED datasheet:

With RGB there is virtually no light emitted at 550nm. We will only perceive that shade of color as a blend of the R+G light, and on a white surface that will look OK since we’ll get a bunch of 500nm and 625nm light hitting our photoreceptors and “see” something in between. However if you had a pigment that only reflected near 550nm, it would appear dark/black when illuminated with this RGB white.

The white plots for cool (BW), natural (NW) and warm (WS) all have a bit of a dip right around 500nm, but it’s still there. I’m not quite sure how these reconcile with the other graphs of white LED light, but my guess is that some of them are showing perceptual intensity, corrected for how we would perceive it.

Anyway, you can change the apparent color temperature by mixing RGB + W addressable LEDs, and maybe hit that R9 a bit more, but it’s not going to fix the spectral range issue.

If you are getting this far into caring about the kind of white light you are making and trying to compete with super high end light and not just game CRI values, then you need a nicer tool IMO, and adding a narrow spike of red might make things worse. IMO most people won’t care this much, and you’d be chasing a super slim niche of a market that is shopping for super high quality white light and cool blinky RGB in the same device.

Not at this time. Yeah, I don’t think they are designed to handle all 4 elements on at the same time. White on RGBW does take less power than R+G+B, but it usually has a similar perceptual brightness for the same value.

Not if you really want the best white light. I don’t know of any high CRI addressables, you’d probably have to make one custom.

This is what I was afraid of. My sensor is not accurate enough.

IMO most people won’t care this much, and you’d be chasing a super slim niche of a market that is shopping for super high quality white light and cool blinky RGB in the same device.

This is my intention though in the long run.

Not if you really want the best white light. I don’t know of any high CRI addressables, you’d probably have to make one custom.

I’ve been looking for a while, best I can find are some strips with separate pixels (RGB and CCT), but pixel density is very far from what i’m looking for to do cool effects with the Pixelblaze (144 LED/M or higher).
I guess a custom strip about 20mm large could fit both LED next to each other and maintain good density, but indeed it seems I have to make a custom one. Any experience with this ? I’m wondering what’s the idea of MOQ for this kind of work ? 100M ? 500 M ? More ?

Do you need the white illumination element to be addressable?

If so, maybe partner with someone like this (assuming it’s not a gimmick), or use their LEDs with addressable driver ICs.

Right now to proceed your best bet may be to modify an output expander. You could customize how you handle RGB, split white out, adjust color, interface with custom strips, etc.

Ah yeah indeed this is absolutely not the same price as my usual white LED.

But i’m veeeery interested by this ! In my opinion light quality makes the difference between a gadget lamp and proper furniture. I think i will order a bit to have a look myself at the difference.

Thank you !