RGB array, why blue so hot?

[VanIsleDSM]

Using some luxeon rebels in an array, 3 red, 3 green, 3 blue. All SMT soldered onto triple mounting MCPCBs and screwed onto a large heatsink. Running 700mA through each colour in series.

I've always thought that the green with only about a 5% efficiency or so, would get the hottest. And the only reason why the appear so bright with such a bad efficiency is because of how sensitive the human eye is to green.

That's not the case at all though, the red and green stay quite cool, while the blue are on fire. Even mounted onto a large heatsink, feeling the dome of the LEDs, they're too hot to touch only a few seconds after coming on.

I'll be turning down the current to the blue, and leaving it the same for the red and green, since the blue is a bit overpowering anyway, but this just seemed odd to me. I'm sure someone here has an explanation though.

I'm reluctant to believe the spec thermal resistance on these rebels, even with the MCPCB, there's no way the dome should be ~100C while the heatsink is completely cool.

Solder has all evenly flowed, and the MCPCBs seem of high quality, they're these ones here: http://www.luxeonstar.com/rebel-star-blank-aluminum-mcpcb-base-triple-led-20mm-p-444.php

But who knows.. without actual specs on the dielectric layer.

 

[jtr1962]

There is nothing wrong. The Rebel has a tiny dome which means the light energy leaving it is highly concentrated. Any object placed on top of the dome will absorb that energy over a small area. Blue is more efficient than red or green, so more energy is leaving the blue. Add to that the fact that your skin absorbs more blue light than red or green. No surprise that putting your finger on the dome causes it to feel hot. It isn't the dome which is hot, but rather your finger from absorbing all that energy. It's similar to measuring air temperature outdoors on a sunny day with a dark probe-you'll get much higher readings than the actual air temperature because the probe absorbs sunlight. Try putting a piece of black electrical tape over a white Rebel. It'll burn right through. Now do the same with white and nothing will happen. Nothing wrong here at all. Try feeling just the MPCB, not the dome. It should be quite cool to the touch.

 

[VanIsleDSM]

Hey thanks jtr, I hadn't quite put that together, but it makes complete sense. Yes everything else is very cool, that's why I was wondering. Thought I was just getting some bad thermal conductivity.

Well I suppose everything is gravy then I've got this thing fading through the whole spectrum nice and slow, and it can light up a whole room, pretty neat. Before this I'd only really played with white LEDs.

Now I just need to find a source for some of these:

http://carclo-optics.co.uk/catalog/documents/Lumileds-01-09.pdf

page 26, triple rebel optic. Can be had in bulk right from carclo, but I'd like to get a few different types of each to play with, can't seem to find anyone supplying them in smaller numbers.

 

[jtr1962]

Well I suppose everything is gravy then I've got this thing fading through the whole spectrum nice and slow, and it can light up a whole room, pretty neat. Before this I'd only really played with white LEDs.

Cool! I few months ago I used red, green, amber, and blue rebels to make a variable color temperature light source. No optics or anything as I was just interested in general lighting (I bounce it off the ceiling to mix colors). Dip switches set the CCT at any of 11 settings from 1900K to 9300K. Fun to play with although whites are a lot easier to work with.

Now I just need to find a source for some of these:

http://carclo-optics.co.uk/catalog/documents/Lumileds-01-09.pdf

page 26, triple rebel optic. Can be had in bulk right from carclo, but I'd like to get a few different types of each to play with, can't seem to find anyone supplying them in smaller numbers.

I'll keep an eye out for those in the course of my web surfing, and let you know if I come across anybody selling them.

 

[Christexan]

Late to the party here, but wanted to chime in that JTR's explanation, although probably somewhat valid, isn't the whole story...

"Heat" emission from a non-contact perspective is in the infrared spectrum, and none of the LEDs project a significant amount of heat, meaning if you hold your finger over any color LED (without contacting it) that has a die temp of "x", they should all be similar. There will be SOME difference, but not "noticeably hot" versus "cool to the touch"....

The larger factor is certainly die temperature (direct conduction heating of the dome)...
You said all are at 700mA (constant current regulated I assume)...
Red and Green LEDs typically have Vfs around 2.0. Blue LEDs are typically around 3.4. So the wattages for them are:
R/G = .7*2=1.4 watts (each)
B = .7*3.4= 2.4.
So the blue at the same 700mA current is consuming a full watt more of power, most of which is being emitted as heat (at least 80% or so). Some goes out the top, some into the MCPCB. 1 watt spreads quickly on the MCPCB (and merges with the heat from the other 2 units, so it's not going to feel a lot hotter as the area quickly spreads the heat), but on a point contact on the dome, even if it's only 40% hotter, that is much more noticeable to the touch. 1 watt isn't much, but concentrated in the tiny area of a rebel LED, is a lot of watts/cm.

 

[SemiMan]

I would agree with you if you were right.....

The Green LED will have almost the exact same forward voltage as the Blue LED as both are based on InGaN technology. There is differences in the band-gap certainly, but the forward voltages are relatively close.

Your other argument about "heat" is flawed as well. By your explanation, I should be able to put my hand in a microwave and not feel any heat. Try explaining that to my hand as it boils and I scream in pain. Good thing I did not experiment with the cat ...

The absorbed radiation of the blue will absolutely cause heating of the skin. Any radiation that is absorbed that is not re-radiated will cause heating of the material at hand in this case your skin/finger.

VanIsleDSM, I made a similar observation when I first put several white Rebel-100s very close together and ran them hard. I was putting my finger over the LEDs to block the light so I did not get spots in front of my eyes. My finger quickly got hot which I assumed was because the LEDs were physically hot. I thought my heatsink/attach was flawed so I got out my temp probe and found the package of the LEDs was quite reasonable and the domes were not that hot at all. Then it dawned on me that those 4 Rebel-100s were putting out almost 1 watt of visible light in an area about 5-6mm on a side or about 30,000 watts/square meter. All of that energy was being absorbed by my finger! No wonder it was starting to cook.

Semiman

Late to the party here, but wanted to chime in that JTR's explanation, although probably somewhat valid, isn't the whole story...

"Heat" emission from a non-contact perspective is in the infrared spectrum, and none of the LEDs project a significant amount of heat, meaning if you hold your finger over any color LED (without contacting it) that has a die temp of "x", they should all be similar. There will be SOME difference, but not "noticeably hot" versus "cool to the touch"....

The larger factor is certainly die temperature (direct conduction heating of the dome)...
You said all are at 700mA (constant current regulated I assume)...
Red and Green LEDs typically have Vfs around 2.0. Blue LEDs are typically around 3.4. So the wattages for them are:
R/G = .7*2=1.4 watts (each)
B = .7*3.4= 2.4.
So the blue at the same 700mA current is consuming a full watt more of power, most of which is being emitted as heat (at least 80% or so). Some goes out the top, some into the MCPCB. 1 watt spreads quickly on the MCPCB (and merges with the heat from the other 2 units, so it's not going to feel a lot hotter as the area quickly spreads the heat), but on a point contact on the dome, even if it's only 40% hotter, that is much more noticeable to the touch. 1 watt isn't much, but concentrated in the tiny area of a rebel LED, is a lot of watts/cm.

 

[IMSabbel]

"Heat" emission from a non-contact perspective is in the infrared spectrum, and none of the LEDs project a significant amount of heat, meaning if you hold your finger over any color LED (without contacting it) that has a die temp of "x", they should all be similar. There will be SOME difference, but not "noticeably hot" versus "cool to the touch"....

I wonder where the strange "only infrared radiation makes things hot" thing comes from. Maybe bad science school curicula.

Energy is energy. No matter what form. In you argument, a red or green laser woulnd burn stuff because its not infrared....

 

[lolzertank]

I wonder where the strange "only infrared radiation makes things hot" thing comes from. Maybe bad science school curicula.

Energy is energy. No matter what form. In you argument, a red or green laser woulnd burn stuff because its not infrared....

I think it comes from "hot things emit infrared," which is not anywhere near completely true, but it works pretty well on Earth. Anything with a temperature above absolute zero is a blackbody radiator, and it just so happens that pretty much everything on Earth is just the right temperature to emit infrared.

 

[evilc66]

Hey thanks jtr, I hadn't quite put that together, but it makes complete sense. Yes everything else is very cool, that's why I was wondering. Thought I was just getting some bad thermal conductivity.

Well I suppose everything is gravy then I've got this thing fading through the whole spectrum nice and slow, and it can light up a whole room, pretty neat. Before this I'd only really played with white LEDs.

Now I just need to find a source for some of these:

http://carclo-optics.co.uk/catalog/documents/Lumileds-01-09.pdf

page 26, triple rebel optic. Can be had in bulk right from carclo, but I'd like to get a few different types of each to play with, can't seem to find anyone supplying them in smaller numbers.

Future Electronics sells all of the 20mm triple lenses from Carclo. The MCPCB from Luxeon star was designed to fit these specifically. These also work with the XP-E/XP-C/XP-G LEDs, as does the MCPCB.

 

[bshanahan14rulz]

Someone on here selling p60 style 3xXP-E dropins is also selling just the triple optics. he has three different kinds of lenses, you could give those a shot..

http://www.candlepowerforums.com/vb/showthread.php?t=233442

link to thread in custom bst.

these should be 26mm overall diameter, I think. I couldn't get the link on your post to work, VanIsle, but most stuff that works for XP-E works for rebels too.

 

[Illum]

Using some luxeon rebels in an array, 3 red, 3 green, 3 blue. All SMT soldered onto triple mounting MCPCBs and screwed onto a large heatsink. Running 700mA through each colour in series.

do take note that W = VI
under the same wattage category different color LEDs utilize different chemicals to emit light, making each color have varying input voltages. If you have worked with 5mms you know that while its not uncommon to forward bias a white T1-3/4 LED with 3.6V it will kill red T1/T1-3/4 LEDs over a short period of time.

I think what happened here is that the blue LED is facing high resistance due to a lower Vin from the other LEDs which permits an increased current drop and its heating up like a resistor. I'm not sure how you you measure the intensity of the blue LED to observe any degradation in output but I'm pretty sure it's impossible to tell by eye.

 

[SemiMan]

He says he is running 700mA into each string...Vf will effect power, but the green will have similar Vf and power.

 

[bshanahan14rulz]

I think as long as the array is still getting more than the total Vf it should work. If a single LED gets at least its Vf worth of voltage across it, it would drop what it uses and pass on the rest to whatever is after it.

 

[VegasF6]

I've always thought that the green with only about a 5% efficiency or so, would get the hottest. And the only reason why the appear so bright with such a bad efficiency is because of how sensitive the human eye is to green.

Blue is more efficient than red or green, so more energy is leaving the blue.

How can I find out how efficent a particular led is, or even in a more general term that blue is more efficient, green is less, white is? etc? Is there a chart? Is that something in the spec sheet per manufacturer? Is it based on lumens per watt, radiant flux or what?

 

[T0RN4D0]

For the poster- although i know you understand now; with LEDs you just have to worry how hot the back is, not the front. The perfect LED would be hot as hell in the front (well the energy going out), and ice cold in the back Out of the front we have the good energy (lets call it light), and at the back we have the bad energy (lets call it heat). In general you should be worried about the LEDs that don't put out any energy in the front, because their pumping it all in the back.

And as it has been noted, light can do 2 things when hitting a surface- either bounce and keep its form, or stay there and transform into heat (typically it does both).

 

[jtr1962]

How can I find out how efficent a particular led is, or even in a more general term that blue is more efficient, green is less, white is? etc? Is there a chart? Is that something in the spec sheet per manufacturer? Is it based on lumens per watt, radiant flux or what?

Unless the manufacturer specifies the radiant flux, as they usually do with royal blue, the only way to figure wall-plug efficiency is to take the lumens per watt, and divide by the luminous efficacy of the emitted spectrum. You can approximate the latter by using the photopic eye response at the center wavelength of the emitted spectrum. It's not perfect ( integrating the response over the entire emitted spectrum would be more accurate ), but it's good for a quick and dirty estimate which gets you within a couple of percent. For example, 470 nm blue is roughly 62 lumens per watt. If a 470 nm blue LED is giving you 27 lumens per watt, then the efficiency is roughly 30/62, or ~48%. A 520 nm green LED has a spectrum with an efficacy of about 500 lm/W. Therefore, an 80 lm/W green LED with a center wavelength of 520 nm would be about 80/500, or 16% efficient.

For white LEDs which are broad spectrum there is obviously no "center wavelength". A good rule of thumb for typical YAG phosphor whites is to use 330 lm/W for the emitted spectrum. Warm white will be a little less, perhaps 300 lm/W, and high-CRI LEDs will be lower still, perhaps 270 lm/W ( not really sure of that number-just an educated guess ). You can then use these numbers to calculate the efficiency. An R5 XP-G might get around 138 lm/W at 350 mA, so its efficiency would be around 138/330, or ~42%.

 

[VegasF6]

Thank you. In GENERAL, can you state then what colors leds are most efficient to least efficient (or would Efficacy be more correct, I dunno) including red yellow blue green and white? For the sake of argument, lets say they are all Luxeon Rebels or any maker that might make all those colors in the same series.

Sounds like green is actually the worst?