Red or red-orange for increasing CRI?

[lolzertank]

Is it better to use red or red-orange LEDs to increase the CRI of a cluster of neutral white LEDs? I will be using Rebels, which are 100 lumen at 350ma whites, 50 lumen at 350ma red-orange and 30 lumen at 350ma reds. They will all be driven at 700ma. Also, what proportion of red/red-orange to white should I use?

Thanks!

 

[jtr1962]

White LEDs are most deficient in deep reds, so red will best fill in this deficient part of the spectrum. As for the amount, you have to experiment a bit but offhand I'd say one red to 4 or 5 whites (assuming all are driven at the same current). Because the addition of red might make the light slightly pink, a small amount of cyan will offset that (and also compensate for the other area where whites are lacking-the valley in the ~500 nm region).

 

[lolzertank]

Thanks! From your advice, I think I'll put 9 white, 2 red and 1 cyan in my cluster.

 

[jeffosborne]

Hey lolzertank, you have my attention! I am preparing to make a photography flood light with 9 neutral white Rebels, too! The 100-lumen parts I got recently from Future Elec. are, unfortunately, the coolest bin code they offer, so I will use the 90-lumen parts, which are a little warmer. I also have my cyan and red Rebels ready to go, 2 and 1, as you mentioned.

I plan to make a tripod-mountable enclosure built around a penium-4 heatsink and fan, and a home-made driver to drive 3 LEDs in series x4, and use a 12 volt power supply or 10.8 volt lithium battery pack. I will be trying various diffusers to complete the light.

Do you have heatsink and driver in mind?

Cheers,
Jeff O.

 

[lolzertank]

I'm going to be using it in a Mag on 2x18650, so I'm just going to use a bunch of AMC7135s and Download's Multi-sink. I wanted to do a triple MC-E/P7, but the P7 only comes in cool white and the MC-E is less efficient than the Rebels for neutral white. At first, I was going to use 12 neutral Rebels, until I thought that since I'm using so many emitters anyways, I could go for some color mixing.

Sorry to hear that you lost the Luxeon lottery. Did you at least get something on the black body line?

 

[HarryN]

Hi, if you are going to this much bother, perhaps there is a more simplistic approach- just use single color LEDs and mix them in the first place.

I am a big believer in 6 color mixing, nominally
450/460nm - strongly blue
488nm - kind of a torquoise / ghost like color
505nm - traffic green
525nm - amber
620nm - approx red / orange
650nm - fairly deep red

488 can be tough to get, so often, people just double up at 505 to get enough green into the light. The other challenge is that red and r/o rebels cannot handle the same current as the others, so you either need to double these up in parallel, or use Lux IIIs.

If you put these into two strings of similar Vf, then a 12 volts source and resistor would work, or a buck puck.

 

[lolzertank]

Hm... doesn't that configuration kind of kill efficiency thanks to the relatively low efficiency of green?

 

[JohnR66]

Cree uses a mixture of red and yellow LEDs in their lighting fixture to obtain a claimed CRI of 90 or more. Of course It depends on the color temp of the main white LED used too.
http://www.creeledlighting.com/lr24.htm

 

[HarryN]

Hm... doesn't that configuration kind of kill efficiency thanks to the relatively low efficiency of green?

With LED lighting, it is always a trade off between good coverage of the spectrum and lumens ratings. LEDs are the most energy efficient (relative to an incan) in single color applications, such as a stop light.

CRI is a rating of how the spectral output of a source is relative to a fairly warm color temp incan bulb. (incan bulb = 100, everything else is worse) When you start getting closer to simulating the very broad spectral output of an incan bulb, the incan starts looking more efficient relative to LEDs than we like to imagine. A good example is the Lumileds Lux I warm white - it is virtually an exact spectral copy of the visible light of an incan, and its efficiency is very low.

In photography, the historic "white color standard" was the sun, so a color temperature of around 5,000 K. The CRI of the sun is actually lower than an incan bulb too IIRC, but most people dont' notice.

With digital cameras, the sensors are often have relatively narrow sensitivity to specific RGB colors, although some also have a cyan signal. This means that for digital photo use, you could potentially get away with a really mediocre RGB setup.

Yes, with RGB, you often have to double up the green.

Lumileds made countless presentations for years at seminars trying to get them to see the advantage of RGB and other multi color LED mixing vs. white, both from an efficiency and lifetime perspective. Pretty much no one listened, so they put a lot of effort into white, just like everyone else. You will notice that most LED backlit TVs are RGB, not back lit with white. There are reasons for that - the TV makers understand the point.

 

[lolzertank]

No, what I was trying to say was that green LEDs just don't produce very many photons. New blue LEDs convert 40% of their energy into photons, but green LEDs are stuck at about 10%. Even if phosphors are only 50% efficient, it's still more efficient to produce green through blue + phosphor. The narrower band of a true green LED obviously doesn't matter to me since I'm trying to get white light. Amber is similar in this respect as well, which is probably why Lumileds made the phosphor converted amber that stomps old amber LEDs despite needing phosphors on top of a blue die.

If green LEDs were as efficient as blue, then I would certainly go with your solution. But until the 250 lumen at 350ma green LEDs come out, I don't think RGB will be a efficient solution.

 

[HarryN]

LEDs specs are always on the move, esp. in the green area. Because of the nature of how LEDs are made, actually some equipment is more suitable for making green LEDs than others. Lumileds "just happens" to own that kind of equipment.

I double checked the spec for rebels, and interesting, if you compare the lumens at 700ma for neutral white, green, cyan, and amber, the posted lumens numbers are all within 10%. This makes at least the RG portion of RGB quite viable.

http://www.philipslumileds.com/products/luxeon/luxeonrebel

I think the amber rebels from blue converted phos have more to do with the exceptional difficulty of making good AlInGaP based amber, then a fundamental problem making good InGaN based green. It also allowed PL to have similar electrical, mechanical, and optical specs for nearly all of the rebels now.

PL is so strong technologically in AlInGaP red that normally I would not expect them to make the red and red/orange from blue + phos, but stranger things can happen. IF they do it, it will have more to do with making the rebel line "all similar" than a real improvement in efficiency, IMHO.

 

[lolzertank]

Here's what 2xTrinity said about green's weakness in RGB lighting. I just realized I'm trying to say pretty much the same thing, just minus the physics. Using 6 colors will have this problem too, since the proportion of green to other colors will have to remain the same.

I realize that your solution will probably have higher CRI than mixing neutral white, cyan and red LEDs, but in a flashlight with limited battery power, efficiency is still pretty important and I don't want to throw all of it away for CRI. Something better than 70 is good enough for me. I'll make a Mag623 if I ever get tired of LEDs.

 

[lolzertank]

Ok, so I built the Mag. Wow, colors outside look awesome. Trees actually look like what they look like in daylight. I'm coming from nasty WC tint Crees and dim/orange 6D Mags. There's a LOT of cyan, I even had to put some "transparent" tape over the optic to diffuse it a little. The neutral white LEDs were SM tint. I tried covering the cyan and red which dulled up the color quite a bit.

 

[jtr1962]

You mentioned earlier that you used 9 white, 2 red, and 1 cyan. How is the white balance? And are you driving all the emitters at the same current?

I think I might make a desk light or reading light based on a similar combination of white, red, and cyan. I've already played around with red, green, blue, and amber rebels to get a variable CCT light source. I use dip switches to set the CCT in 9 steps from 1900K to 9300K. Color rendering is fantastic compared to white emitters, but of course efficiency isn't too great.

 

[jeffosborne]

Hey lolzertank, we would love to see a photo of your modified light. Also, beam shots are good!

Thanks,

Jeff O.

 

[lolzertank]

The red is being driven slightly lower than the rest, around 500ma. The rest of the emitters are at 700ma. There's a lot of cyan, much more than is needed. Also, the triple optics aren't perfectly aligned into one hotspot causing one corner to be very green. You probably won't have that problem if you aren't using narrow optics like me. Covering the optic of the cyan with some tape to diffuse it solved both problems pretty well. Still a bit cyan on the edges though, not a big deal to me. I'm not really a white wall hunter.

 

[Gryloc]

Just out of curiosity, how much does the increase in CRI affect efficiency so much when adding red or cyan emitters into the mix? Lumens are lumens, so when you have so many lumens in white, then you add more colored light, the total output does increase.

Lets say we have a cluster of 6 emitters using the best that the Rebel line offers. This is assuming that the colors mix to provide a slightly warmer white than with the bare cool white emitters alone. I have not had a chance to mix the colors in the below example, so I am not sure if the ratios will work out okay to get a balanced neutral or warm colored light.

In series at 350mA:
4 White     100lm    3.15V    1.103W    90.66lm/W
1 Cyan       70lm    3.15V    1.103W    63.46lm/W
1 Red        40lm    2.90V    1.015W    39.41lm/W
--------------------------------------------------
6 Total     510lm   18.65V    6.530W    78.10lm/W

In series at 700mA:
4 White     180lm    3.40V    2.380W    75.63lm/W
1 Cyan      130lm    3.40V    2.380W    54.62lm/W
1 Red        85lm    3.60V    2.520W    33.73lm/W
--------------------------------------------------
6 Total     935lm    20.6V    14.42W    64.84lm/W

Sure, efficiency takes a direct hit, but in these cases, the resultant efficiency does not seem that bad for a small improvement in color rendering. Even Lumiled's new CCT/CRI binned Rebels with higher CRIs have efficiencies in this range at only 350mA. I am not sure of the efficiencies of the high CRI Seoul emitters out there.

My quad TFFC K2 0200 Mag produces 900 lumens (4 * 1.125 * 200lm @1.2A) and consumes 16.32W (4 * 3.4V * 1.2A). My lumen figure is based off of Lumiled's datasheets, and the forward voltage of my K2 emitters are based on actual measurements. These relatively modern, and slightly warmer tinted cool white emitters are operating at 55.15lm/W to produce what appears to be a splendidly bright beam. Even though 0200 binned TFFC K2s are not that amazing in terms of performance to many here (I know there are better performing parts out there), in my case, they do perform very nicely. However, the efficiency is still far below that of the 6 above Rebels operating at lower current.

Can you guys please produce a simple chart showing known good ratios of lumens of each color required to get a few known CCT whites using white, red, and cyan, or maybe other colors? That would be extremely useful and helpful in future projects! From a ratio of required lumens of each color, we can determine the proper currents from each of our emitters to get a white color on that is close to blackbody line. Thanks!

-Tony

 

[LEDAdd1ct]

Beamshot?

 

[lolzertank]

The camera won't be able to show the CRI increase very well thanks to only having RGB filters on the sensor. Not to mention white balance totally messing everything up and the necessity for a calibrated screen to accurately show colors. I also don't have anything good to compare it to. Well, unless you want to see this ~1900 lumen monster against a 10 lumen Fenix E01 and its angry blue/purple tint. Now I wish I didn't lose that Cree XR-E DIY kit with all it's WC tint glori/ugliness...

The beam is extremely floody though, about 40 degrees from edge to edge. There's spill, but it's really not that useful since the hotspot is so wide. It doesn't actually look THAT bright until you think how wide the beam is. It has about the same throw as a TK10 after all.

The Fedex truck delivered the LEDs to me the day before Lumileds decided to put the high CRI Rebels on their website. My sig became real, just with CRI instead of flux.

Gryloc, your mixes have way too much cyan. 1/4 of the amount would be enough, or else your light will start looking like a traffic light. That should help efficiency a little. You could also use the neutral white emitters which would give a lot more red and green.

I think the neutral white high CRI P4s are 72 lumens at 350ma. They have a higher Vf than the Rebels too, 3.25V at 350ma. That's 68 lumens/W. The high CRI Nichia 083 is 60 lumens at 300ma and its Vf is 3.3V, which makes it just about 61 lumens/W.

 

[blasterman]

Color rendering is fantastic compared to white emitters, but of course efficiency isn't too great.

Yeah...efficiency takes a hit. I was going to try an experiment building a mixed light bar along the same route with the goal to achieve +90 CRI. However, common sense keeps intervening and begs the question why not use high CRI Seouls to begin with.

I've also noted I can't break my addiction to mixing neutral white and warm white emitters for fixed lighting because I swear CRI is better with the emitters combined.

Also looks like Cree is mixing amber and neutral white emitters with their +90CRI ceiling fixtures.

One approach I've taken to this is looking at the spectral graphs for high CRI emitters like the high CRI Seoul and typical neutral/warm-white emitters. You can kind of get an idea of what colors are missing looking at the difference in the graphs.