LS Lux vs Colour question

Anyone know the answer?

I have tested 2X TWAK 3W LS LEDs, both from the same batch and both in the same host and setup. The first gave a near white colour, but leaning towards blue. The Lux at 1m was just over 12,000 at 2A. The second I was more blue but gave 13,700 lux. This had me thinking....

Assuming the same LED die efficiency (a big assumption), is it possible that knowing the LED is blue and the phosphor being a green/yellow adds colour to the spectrum making it appear white, that the whiter (or greener) the LED is means the phosphor is thicker and therefore blocks more light from the LED and conversely, that the more blue a LED that the phosphor is thinner resulting in more light from the LED getting through the phosphor and making the lumen output higher?

Perhaps Lux isn't everything, nor is colour.

Chris

The variation in light output is just due to random variation in die efficiency. You really can't make any generalizations based on two samples. I have 5 Q3J 1W Luxeon stars that I bought on eBay. Four of them are appear identical in brightness and color, while the fifth is slightly less blue and slightly dimmer. While this may appear to support your theory, I've seen plenty of R2 1 watters on eBay but no R3's. If anything, the reverse should be true. Bottom line, I'm not 100% sure if we can draw any generalizations because there is still a wide variation in the optical output of the blue dies from batch to batch, and I tend to think that affects output more than the thickness of the phosphor layer. I'm not discounting your theory 100%, but until I observe the same pattern on more samples, especially samples from the exact same batch of blue dies, I remain unconvinced.

As an aside, why on earth are you running a part rated at an absolute maximum of 1 A at 2A? For starters, as a valid test of your theory you would need to run the parts within their ratings. If you run them at higher currents anything can happen colorwise. Secondly, efficiency drops at higher currents, and looking at the data sheet there isn't a whole lot to be gained output-wise going from even 700 mA to 1A, much less to 2A. You're probably gaining 50% in light output relative to what you would get at 700 mA, using about 3.5X the power, and shortening the lifetime from 50,000 hours to probably a few hundred at best(assuming the thing doesn't get pissed off and die instantly due to that much current). What's the point? You can get twice the light output that you're getting now from overdriving that 3-watter, and even use slightly less power than you are now, simply by running three 3-watters in either series or parallel at their rated currents of 700 mA. This is also true to some extent for anyone overdriving 1-watters, although at least up to about 700 mA the 1-watters don't have that bad of a dropoff in efficiency. Past about 700 mA for either part, you reach a point of diminishing returns, and I wouldn't drive anything past it's absolute maximum anyway(i.e. the 1-watters go at 350 mA max regardless of light output at higher currents). As an electronics engineer I know there is a reason for absolute maximum ratings. They should be respected or you'll have lots of expensive parts blowing out.

jtr... Good points. However, you are coming from the school of thought that specs should be observed and that is that. Sure, if we were making lights for the masses, I'd keep it at its max ratings. Definately. Here at CPF, we have no constraints. We push the envelope MUCH harder and faster. No sooner does something new come out, it is pushed and pushed. Sometimes at the detriment of the parts we use. /ubbthreads/images/graemlins/icon3.gif That's just part of the experience. Would you considering running a 5w LS (max rating 700mA) at 1.5A? I do.

Oh, by the way, welcome to the CPF /ubbthreads/images/graemlins/grin.gif /ubbthreads/images/graemlins/grin.gif

350ma is the maximum rating for a white star in open air. With a flashlight that provides good heat sinking, you can certainly exceed it.

Also, although you are quite correct that running multiple luxeons at lower power provides more light, this isn't usually practical for use in a flashlight. The tri-star accomplishes this (with a special Fraen tri-lens), but it requires a large head.

[ QUOTE ]
hotbeam said:
jtr... Good points. However, you are coming from the school of thought that specs should be observed and that is that. Sure, if we were making lights for the masses, I'd keep it at its max ratings. Definately. Here at CPF, we have no constraints. We push the envelope MUCH harder and faster. No sooner does something new come out, it is pushed and pushed. Sometimes at the detriment of the parts we use. /ubbthreads/images/graemlins/icon3.gif That's just part of the experience. Would you considering running a 5w LS (max rating 700mA) at 1.5A? I do.


[/ QUOTE ]
There's nothing wrong with pushing the envelope and a little destructive testing. I hope I didn't give the impression otherwise. While I personally can't afford to destructively test $10 parts on a regular basis, I'm certainly interested in reading about it here, and seeing exactly how far these parts can be pushed.

With regard to absolute maximum ratings, yes, I would observe them also on anything sold commercially to a general audience. I'm quite aware that you exceed those ratings on items you sell, but those who purchase your products are well aware of this, and you do seem to take every reasonable precaution. Again, for my own personal use or a small audience, I might exceed absolute maximums provided those who bought my products knew the ramifications of this.

I personally have a couple of thoughts on exactly how much an LED like the Luxeon star can be overdriven. There are two failure modes-the die can cook from overheating, or you can exceed the limits of the wire bonds even without overheating the die. For the 1-watter I would probably feel fine going to 500 or 600 mA, provided the heatsinking is adequate to keep the LS heatsink at 40°C or less. At 600 mA and with a die to heatsink impedance of 20°C/W this would mean a die temperature of ~85°C. This is certainly within limits and you have some safety margin. The bond wires can withstand 350mA for 50,000 hours(at least). They can certainly take maybe 600 mA for the much shorter lifetime the LED would have at that current level. In something like a flashlight the much shorter lifetime at 600 mA is irrelevant. I'm not so sure how I would feel going much past about 600 mA(maybe 700 mA tops). At that point I think we might run into issues with the bond wires.

The 3-watter is a different animal, and here I think bond wires aren't the issue for any current level because the thermal limits would be reached long before the bond wires were unduly stressed. Drawing an analogy similar to what I did for the 1-watter, I would say we're safe keeping the current level for the 3-watter under 1.5A at least with regard to the bond wire issue. However, we run into heat-related problems long before that. Once again, say we manage to keep the Luxeon III heatsink at 40°C with good external heatsinking. At a die to heatsink impedance of 17°C/W the die is already at 106°C running at the absolute maximum rated current of 1A. If you go to 1.5A(and say 4.2V), the die temp is now 147°C. At 2A and 4.5V(i.e. 9W) you're up to 193°C. The absolute maximum die temperature is 120°C. Beyond that you're literally cooking the die to death in a very short time. This is why I felt so strongly about overdriving a L3 at 2A. I would feel fine driving one at maybe 1.2 or 1.3 A with a really good heat sink, or going to 1.5A(or even a bit more) with active cooling. However, at 2A you need to keep the star's heatsink at -33°C or less not to exceed thermal limits, and I feel the bond wires are being unduly stressed at that current level regardless. Of course, for a second or two just to get a Lux reading I might do it out of curiosity. /ubbthreads/images/graemlins/wink.gif

Regarding the 5-watter, 1.5A is probably a bit much for the bond wires, and certainly well beyond the thermal limits of the package. Even driven to specs, the 5-watter has a very short lifetime(presumably due to heat-related issues). Of course, for a second or two I have no problems driving at 5-watter at that level. It does appear Lumileds solved the thermal problems with the LS3, which can be driven at 1A and ~3.9W yet still have a 20,000 hour lifetime(or 50,000 hours at 700 mA).

One reason I've raised eyebrows at overdriving LEDs is because past a point there just isn't much gain in lumen output. If the lumen gain were nearly linear, I might at least see some reason for it. Nothing wrong with having an LS3 driven at 700 mA normally, and maybe a 1.1 or 1.2A "turbo" mode. Beyond that I don't think there's much more light to be found. Overdriving bulbs, on the other hand, makes more sense provided you're aware of the greatly shortened lifetime. Bulbs actually get much more efficient(and whiter) when they're driven harder, and these are both desireable characteristics.

[ QUOTE ]
Oh, by the way, welcome to the CPF /ubbthreads/images/graemlins/grin.gif /ubbthreads/images/graemlins/grin.gif

[/ QUOTE ]

Thank you very much. /ubbthreads/images/graemlins/smile.gif I appreciate it. I've been lurking here the past few weeks on and off, and finally a thread came up that I just had to post in(that other thread regarding Moore's Law and LEDs). The topics of most interest to me generally are similar to that thread and this one. While I don't see myself ending up with a huge flashlight collection, I do like to tinker with LEDs from time to time. I already modified two bike lights with some LEDs from ChiWing on eBay. Good LEDs for the price-maybe a bit less efficient than Nichia, but nice even beams and good quality. The one I did with 20 8,000 mcd LEDs gives a really nice light on a dark street, although here in NYC you barely notice the beam on most streets due to the streetlights. I have yet to do anything productive with my 5 1-watt LS's, but those were something I wanted to play with ever since I learned about them. /ubbthreads/images/graemlins/grin.gif

jtr1962,

A few questions to answer...

True re "If anything, the reverse should be true. Bottom line, I'm not 100% sure if we can draw any generalizations because there is still a wide variation in the optical output of the blue dies from batch to batch".

These were from the same batch. I still believe however that it is a wild assumption that the lumen o/p were identical hence asking the question what's the go.

Re overdriving,

I too came from the same camp as yourself. I have since "seen the light". Simply, 50,000 hours is a rediculous log run-time for a torch for most users. With the speed LED technology is increasing, I would be glad after a "few hundred" hours of use to replace a $10 LED for one with much higher spec. Seen the price and run-time of high spec incandescents lately of these high-spec incandescents? Keep in mind, the LED is spec'd at a reduced light output rating, not non-functional.

So why do it?,

"I'm certainly interested in reading about it here, and seeing exactly how far these parts can be pushed."

....that's why. So am I.

Re Thermals,

"At a die to heatsink impedance of 17°C/W".....is also a wild assumption. Personally I believe the heatsink combo of emitter (not star), arctic silver epoxy, hotlips and a Mag D tube and head is much less, but without drilling and potentially destructive test methods being employed, this remains an unknown. Keep in mind, the LS datasheet has junction to case (the body of the emitter assembly) spec'd at 1.3C/W. 17C/W sounds very high for this mod.

And to let you know,

re "The 3-watter is a different animal, and here I think bond wires aren't the issue for any current level because the thermal limits would be reached long before the bond wires were unduly stressed. Drawing an analogy similar to what I did for the 1-watter, I would say we're safe keeping the current level for the 3-watter under 1.5A at least with regard to the bond wire issue."

The 1W star is not well thermally coupled to the heatsink. It is via PWB laminate. I directly connected the actual emitter of the LIII to the heatsink via acrctic silver. It is not a poorly thermally conductive assembly like the 1W star.

Lets see what goes huh? You wanted to see how far these can go, so let's do it.

Also, welcome /ubbthreads/images/graemlins/smile.gif

Chris

[ QUOTE ]
Burnt_Retinas said:
These were from the same batch. I still believe however that it is a wild assumption that the lumen o/p were identical hence asking the question what's the go.


[/ QUOTE ]
Now that's a little piece of info that you didn't mention in your original post which changes things enormously. I didn't discount your theory 100%. In fact, I still think it is entirely possible. I would just like to see some sort of pattern on many units from the same batch. Perhaps someone who buys these by the hundreds might be able to find the pattern. Manufacture of white LEDs is still more of an art than a science, at least regarding consistency of color and output. Until that is fixed, LEDs will not be able to penetrate the general lighting market even though they're the best choice by far for small lighting needs.

[ QUOTE ]

I too came from the same camp as yourself. I have since "seen the light". Simply, 50,000 hours is a rediculous log run-time for a torch for most users. With the speed LED technology is increasing, I would be glad after a "few hundred" hours of use to replace a $10 LED for one with much higher spec.


[/ QUOTE ]
I can't argue with this line of reasoning. In fact, I even said myself that in a torch the shortened lifetime caused by overdriving is entirely irrelevant, and by the time the LED goes, you will likely be able to replace it with something three times as efficient. Once LED efficiency plateaus out, however, then I think lifetime, even for torches, will be a bit more relevant. Regarding the 50,000 hours being ridiculous, remember that Lumileds is after the general lighting market where the (currently)more expensive LED must have something to offer compared to other alternatives. Right now except for the very best specialized incandescants no bulb can touch the ~30lm/W of white LEDs, and no other form of lighting, period, can match the lifetime. I would even say that I would like to see LED lifetime go up to 200,000 to 300,000 hours so you could sell LED fixtures for the general lighting market where the light source never needs replacing.

[ QUOTE ]

The 1W star is not well thermally coupled to the heatsink. It is via PWB laminate. I directly connected the actual emitter of the LIII to the heatsink via acrctic silver. It is not a poorly thermally conductive assembly like the 1W star.


[/ QUOTE ]
Again, a little piece of info that I didn't have earlier. The fact that you were able to put ~9W to the emitter and not have it blow up in your face instantly is telling me that the thermal impedance is probably under 10°C/W. This can be measured nondestuctively by using an infrared thermometer or infrared imaging. I don't have this equipment, and I doubt anyone else here does. However, you would apply a known power to the LED while keeping the heatsink at a known temperature and measure the die temperature. I for one would really be interested in seeing the results of such an experiment. The L3 datsheet specs the junction-to-case thermal impedance at 17°C/W. This consists of the junction-to-emitter and emitter-to-case impedances in series. I would say that the junction-to-emitter impedance is the limiting factor, and you can get the emitter-to-case impedance down to very low values using thermally conductive adhesives. In fact, using plain old thermal grease and carefully machined surfaces I regularly get ~0.01 to 0.02°C/W impedances with 40mm square thermoelectric modules. Scaling down to something like a LS emitter, I think we could get emitter-to-case impedance down to ~1.5°C/W or less. As I said, I think the limiting factor will be the junction-to-emitter impedance, and this is not on the datasheet.

[ QUOTE ]

Lets see what goes huh? You wanted to see how far these can go, so let's do it.

Also, welcome /ubbthreads/images/graemlins/smile.gif


[/ QUOTE ]
Thank you, and by all means continue your experiments. /ubbthreads/images/graemlins/smile.gif I'm really curious as to how far these things can be pushed myself even though this isn't something I can afford to do personally. BTW, is there a whole lot of increase in light output above 1A? I surmised that there wasn't based on extrapolating the datasheet curve, but I'm open to being wrong here, in which case overdriving makes a bit more sense than before.

As an aside, I found that experiment fascinating where someone here cooled a LS via a thermoelectric module. I may do that with one of my LS's as it certainly won't hurt it. Using liquid heatsinks and with the tap water starting to get fairly cold now I can obtain cold plate temperatures of about -40°C or less. If I do anything like that I'll post the results regarding increases in light output. /ubbthreads/images/graemlins/grin.gif

[ QUOTE ]
Burnt_Retinas said:
Yclo,

Yeah, but the MR overdrives the 5W. I'm making the assumption 12W would be rated power, yet capable of overdriving too. Perhaps to 20W?

LIII Lux vs current I got today suggetsts overdriving the new 'thermal solution' die is a real possibility. For the record:

Setup = TWAK LIII, Hotlips HS, AS epoxy LED to Hotlips, Mag D host, UCL, distance to light meter = 1m.

Results, current then lux:

0.25 = 3480
0.5 = 5670
0.75 = 7480
1.0 = 8820
1.25 = 9970
1.5 = 10950
1.75 = 11700
2.0 = 12310

I also noticed lux dropped as temp/time rose doing the test. I allowed things to cool for 30 seconds or so then threw in the following:

2.25 = 13600
2.5 = 13700

As can be seen, lux vs current drops off as we approach 2A then there's not much to be gained after that when hot (cold gives outstanding results). Personally I'd stick with 2A max for the LIII. This assumes a premium heatsinking solution as was used.

Now a 12W thermally enhanced die config at 20W?

Chris


[/ QUOTE ]

[ QUOTE ]
yclo said:
My apologies Chris, I thought you meant 12W as in 12W and not overdriven 12W if you know what I mean.

Anyway, here's a plot of the data above for those that like visual stuff (like myself).



Now a rated 12W would be cool! (or hot..)

-YC

[/ QUOTE ]

Copied from this post here...

Thanks for the info, hotbeam. Very interesting. Normalizing everything to the light output at 700 mA (~7000 lux), I see that we get about 76% more output at 2A relative to 700 mA and 40% more relative to 1A. Not quite as bad as I had initially thought. I suppose if you're mainly interested in eeking out every bit of possible light then going to 2A makes sense. Of course, going by the same figures efficiency at 2A is about half what it is at 700 mA, so you are making mostly heat at 2A. It seems the point of diminishing returns starts at around 1.5A. Another thing I thought interesting was the figures at 2.25A and 2.5A before everything had a chance to heat up. First of all, I'm amazed that the L3 could take that level of current, even for a few seconds. Second, this tells me if the thermal path from the die to the emitter were better, coupled with really good heatsinking, we might get quite a bit more light out of these things at 2A. Given the size of the die and the latest thermal adhesives, I'm thinking something less than 1°C/W thermal resistance from die to emitter might be possible. I've already shown that very low values of emitter to case(i.e. heatsink) are possible. Get that total thermal resistance from die to case down to 1 or 2°C/W and we could likely see 30% or 40% more light at 2A than we currently do. Remember that a lot of the reason the curve isn't more linear is due to the light output dropping as the die gets hot. Regarding bond wires, it is important to remember that there are heating losses here as well, and they are proportional to current squared(hence my reason for going ballistic when I first read about driving an L3 at 2A).

Very interesting experiments, and this is seeming to suggest a path Lumileds should take in the future-better die to emitter thermal bonding and heavier bond wires. Both will contribute to enhanced efficiency and reliability at higher current levels. This should start to get really fascinating in a few years when quantum efficiency starts to hit 50% or better as you'll be needing to deal with less and less heat relative to the power input. Overdriving will probably be less a lesson in diminishing returns at that point.

BTW, those things must be really painful to look at. My Q3 1-watters driven within specs produce 31-40 lumens according to the bin number info, have no collimating lens, and are already painfully bright to stare at. If and when we start hitting 1000 lumens in a package(it'll happen within a decade), there exists the definite possibility of causing permanent eye damage staring at one of these.

jtr1962,

It's nice to see you are thinking about this and a tinkering sort. The more eye's ears and minds the better.

Re cooling LED's, if you do have some results, please post. Though it is impractical to include such in a torch mod (perhaps?), it will answer how much effect heat has on efficiency. I suspect a lot with these. I've seen the lux readings drop with all LS LED's right from switch-on. It's one of the reasons I personally go for a decent heatsinking solution and host.

Now you've got me rummaging through my junk boxes for those peltier's I have hidden somewhere.....

Hotbeam,

Thanks for finding the info. Would have been another late night otherwise.

Chris

Bump to show the Lux vs current graph for a Luxeon III