Realistic simulations of luxeon output

As much as the term "realistic simulation" is an oxymoron /ubbthreads/images/graemlins/tongue.gif, I've attempted to model the output of various luxeons.

Lumileds provides output ratings for luxeons, but based on the quite unrealistic situation with a junction temperature of 25C. Unless you plan on immersing your luxeon in ice water, or running it at a 1% duty cycle, the junction temperature will be higher than 25C. Thus, the output you can expect will be much different than the data sheet specifies, depending on your application.

Lumileds also provides output degridation data for various junction temperatures. This is one point that has prevented me from posting this sooner is a conflict in information that I've come across, between lumileds datasheets, and data given by lumileds representatives in technical presentations.

The degridation curves from the datasheets show a linear decrease for white/blue/green luxeons (1, 3, 5 W) starting at 0% at 25C, and decreasing to 70% output at 100C. The curves for red/orange/amber show 0% at 25C, and decreasing to 30% at 100C (exponential).

Now, a presentation I've run across, dated November 2003, (available at http://www.netl.doe.gov/ssl/PDFs/Krames.pdf ) that shows the white/blue/green output at 90%+ at 100C, and 70% for red at 100C, 38% for amber at 100C. This could reflect an improvment in the high temperature performance of luxeons, or could reflect best, in-lab results.

As a result, and the extremely limiting nature of the datasheet numbers (modeling based on the data sheet numbers gives performance that has easily been exceeded by CPF members - numbers that the model says can never be achieved have been by several modders), I've decided to bump up the numbers from the datasheet degridation curves slightly, but not upto the numbers in the presentation.

So, with the output degridation model, samples of Vf curves, and a curve of theoretical output vs. input current, I've constructed some curves that show the relative expected output at different current levels, based on different Junction-Ambient thermal paths. All curves modeled with an ambient temperature of 25C.

Remember, the thermal resistances each curve represents is the junction-ambient thermal resistance. To find the thermal resistance of the heat sinking source (head sink, light body, hand, etc.) you must subtract the junction-slug thermal resistance of the emitter. For example, the 1W has a 25C/W curve modeled. To find the thermal resistance of the heat sink, subtract the 15C/W junction-slug resistance of the emitter, to get 10C/W. Thus, if you wanted your emitter to behave like the 25C/W curve, your heatsink must have a maximum thermal resistance to ambient of 10C/W. The lowest resistance curves are provided as an example of the best-possible performance - that is, if the emitter were bonded directly to an "infinite" heat sink.

The "relative to rated output" scale means that the output has been normalized to the number of lumens. Thus, if a luxeon is rated to output 35 lumens at rated current, then 1.0 on the scale = 35 lumens...1.2=42 lumens, etc. This lets you see relative gains from various levels of overdrive.

Rated current used:
1W White - 0.35A
1W red - 0.35A
3W white - 0.7A
5W whote - 0.7A



First up, the venerable 1W white:
The 15C/W curve represents the best possible thermal path from the emitter to ambient. The vertical red lines represent the current at which the junction temperature is exceeded for the curve that it stops at.


This shows the junction temperature. The thick red line is the maximum junction temperature allowed. The vertical line traces down to the current axis, so you can see the current at which the junction temperature for that thermal path is exceeded.



Next, a 1W Red:
The 18C/W curve represents the best possible thermal path from the emitter to ambient. Note that I only modeled to 1A for the red, since I don't have any Vf data beyond 1A (translation - I haven't intentioanlly destroyed one yet /ubbthreads/images/graemlins/tongue.gif )





Next, 3W white:
The 13C/W curve represents the best possible thermal path from the emitter to ambient.




You'll notice that the 3W essentially behaves the same as the 1W, except for the best-case curve, because of the L3's improved junction-slug thermal path. Most of the 1/3W curves are modeled with the same junct-ambient thermal resistance.

Finally, the 5W:
Best thermal path is 8C/W.





The one interesting thing to note about the whites is that the better thermal path, the higher the output will be, the more current can be pumped in before exceeding the maximum junction temperature, and the peak output happens at higher currents (none of those observations should be surprising).

The red curve is interesting, mostly because of the exponential dropoff in output vs. temperature. The peak for red/amber/r-o luxeons is near 500-600mA for most conditions.

Okay, so that was a lot of information. I'm not sure if I presented this in the best way possible. I would really like to present what we can realistically expect as far as real-world performance of luxeons at various current levels. One thing that comes to mind right away is to model the thermal paths as having equal heat sink devices attached. So instead of all being modeled at 30C/W, instead being modeled at R(j-s)+15C/W. Sound good?

What do you think?

"What do you think?"

I think it's very interesting. I also think it's a good idea to stay away from boiling water temperatures.....

Excellent work, thanks for sharing.

Doug Owen

very nice work indeed.
thanks for sharing this great data to us. /ubbthreads/images/graemlins/thumbsup.gif

Some things I forgot to mention:

These models were done with specific Vf luxeons. I tried modeling adjusting the Vf slightly either way. The result was that a lower Vf results in a slightly lower junction temperature, and slightly higher output at a given current, and a slightly higher allowed peak current before reaching Tj(max). How much of a difference? 1 Vf bin difference (i.e. J->K) won't make much of a difference.


You can use these curves to estimate performance in your light. The best-case curve can be used as an estimate of initial performance of a cold light just switched on (since the heatsink/light body hasn't heated up yet). I would guess that a MagLite body has a steady-state thermal resistance of around 5C/W. Thus, a 1W in a mag would follow a curve somewhere between the 18C/W and 25C/W curves. A 5W in a mag would follow a curve between the 11C/W and 15C/W curves - again, steady state. Due to the large thermal mass of a maglite, the best-case curves could be used as a guide for the first few minutes until the light begins to heat up.

Some trends that are interesting to note from the above graphs: A better thermal path results in higher output at a given current, lower junction temperature at a given current, and a higher current level at which the peak output is achieved. The peak output is a result of the theoretical output vs. current curve balancing out with the output derating vs. junction temperature curve. It may be a good idea to consider the curve past the red line (the red zone of death?) as somewhat unknown, since Lumileds doesn't provide derating data beyond 120C. It could be that much greater degridation begins to occur beyond the maximum junction temperature.

Also, as many have said here before, this goes to show just how little the AlInGaP (red/orange/amber) Luxeons benefit from overdriving. In a modest light body, the output begins to decrease beyond 500mA of current. From reading tech presentations, the red/orange/ambers are actually the most efficient LED materials, and can take extremely high current densities - provided the junction can be kept cold. At high temperatures, the output drops off dramatically.

[ QUOTE ]
evan9162 said:
You can use these curves to estimate performance in your light. The best-case curve can be used as an estimate of initial performance of a cold light just switched on (since the heatsink/light body hasn't heated up yet). I would guess that a MagLite body has a steady-state thermal resistance of around 5C/W. Thus, a 1W in a mag would follow a curve somewhere between the 18C/W and 25C/W curves. A 5W in a mag would follow a curve between the 11C/W and 15C/W curves - again, steady state. Due to the large thermal mass of a maglite, the best-case curves could be used as a guide for the first few minutes until the light begins to heat up.


[/ QUOTE ]

Okay, which Mag body are you referring to?

Care to make a guess on the small ARC4 body for C/W from the LED die to ambient air? How about if held in the hand?

D-Cell mag bodies.

You can find the thermal resistance yourself. Simply set the Arc4's brightness level so that the LED is being given 1W of power...I think there are threads in the Arc flashlight forum that say what power level this is (I don't know this personally, nor do I have an Arc LS of any kind).

Anyways, turn on the light, and let it sit for a while (20-30 minutes?). Then measure the temperature of the flashlight body (outside is fine for a rough estimate, inside the body would be better if possible). Measure the temperature of the ambient air. Find the difference, convert to centigrade if you measured in degrees F, and that's the thermal resistance in C/W of the flashlight body. I don't know what the thermal resistance from slug to body of the Arc4 is - I think it's pretty low (a few C/W). Add that to the 15C/W junction-slug resistance of the 1W emitter, and you have the total thermal resistance that you can look up in the above charts.

Another idea of how to use this:

Lets say you want to achieve a certian light output level. You have a luxeon of a certian bin (a V3T 5W), and want to know if you can hit 150 lumens with it, and what current to run it at.

Let's also say that it's in a 2D mag body, so you'll get about 5C/W steady state slug-ambient thermal resistance.

Because the mag body has a high thermal mass (it takes a few minutes to warm up the head and front of the battery tube), you can look at the 8C/W curve at first to see what the first few minutes of behavior will look like. Afterwords, somewhere between 11C/W and 15C/W will describe the thermal properties of the Mag body. If you want to be very careful, choose the 15C/W curve.

A V-rated 5W will output between 113-147 lumens (at 25C Tj). Let's go halvsies to get an average V-bin, at 130 lumens.

So, on the 5W chart, 1 will represent 130 lumens. Divide 150 by 130 to get 1.15. Find where 1.15 is on the 5W chart. Notice that on the 8C/W curve, we can hit 1.15 (or 150 lumens) at about 1.1A. So, you can hit 150 lumens with an average V-bin 5W for the first few minutes of operation.

However, once the light body heats up, and your thermal resistance is around 13C/W, you can't hit 150 lumens at all. Due to the flashlight body heating up (and thus, the LED junction heating up), you'll hit about 130 lumens...And you will begin to approach the maximum junction temperature.

evan9162,

WOW, what a lot of information. Just answered in one stroke many of the questions I've been saving up.

Thanks,
CY

Darin:
Great work and great presentation. This thread should be required reading for every modder who aspires to make the eyes bug out of Luxeons.

Due to an update in the lumileds datasheet, I will be posting new curves for the 1W white. Lumileds now rates the 1W white to a maximum Tj of 135C, up from 120C.

I have updated the 1W output and Tj charts to reflect the new information presented in the datasheets. 1W luxeons are now specced to a maximum junction temperature of 135C.

[ QUOTE ]
evan9162 said:
D-Cell mag bodies.

You can find the thermal resistance yourself. Simply set the Arc4's brightness level so that the LED is being given 1W of power...I think there are threads in the Arc flashlight forum that say what power level this is (I don't know this personally, nor do I have an Arc LS of any kind).

Anyways, turn on the light, and let it sit for a while (20-30 minutes?). Then measure the temperature of the flashlight body (outside is fine for a rough estimate, inside the body would be better if possible). Measure the temperature of the ambient air. Find the difference, convert to centigrade if you measured in degrees F, and that's the thermal resistance in C/W of the flashlight body. I don't know what the thermal resistance from slug to body of the Arc4 is - I think it's pretty low (a few C/W). Add that to the 15C/W junction-slug resistance of the 1W emitter, and you have the total thermal resistance that you can look up in the above charts.

[/ QUOTE ]

Hey evan9162

You ever measure the Mag 2AA body for thermal resistance? Care to take a stab?

Nope, I haven't. I don't even have an LED mod in one, so I haven't got a clue there.

Hey, evan9162, would you believe your thread is still relevant today, or more so, than ever before?
(higher power parts...)

Yes indeed. Now that I have a light meter, I need to do some real comparisons between these theoretical curves and what actually happens in real life. I can tell you now, that I think these theoretical curves are actually conservative, and that the devices perform a little better than thermal losses would seem to dictate.

That can point to several possibilities

A) thermal resistance varies from the stated thermal resistance
B) thermal performance is better than the given graphs (think of the graphs as a worst case guaranteed value)
C) higher efficiency LEDs mean that a more significant portion of the dissipated power in the device is going out as light, and that needs to be factored in when calculating Tj.

Or all of the above. B) can be tested for by varying the heat sink temperature. A) and C) are more difficult to determine.

Here are some measurements I did for one of the old LumiLEDs Luxeons:


Here is a plot from a newer one: