Lux3 current to lumen formula

[andrewwynn]

update: updated the calculated flux to include loss due to heating.

Possibly useful information for modders:

I created a formula to convert current to lumens for this project, based on the lumileds specifications..

Lumen=((1-(0.7-i)^2)*0.6+((i-0.7)+0.4))*60

To compensate for the loss due to heating.. i made the assumption that as in my light.. the heatsink temp is only about 5C over ambient... if that's the case... the formula changes to:

AdjLumen = Lumen * (1 - .2i) .. looks really simple but yikes took a while to reverse engineer.

i is current in amps (not milliamps) and it's really easy to replicate this formula into an excel spreadsheet and help figure out what current levels you'd need to get a certain amount of brightness.. less trial and error.

you can use a 'goal seek' to figure out for example what current level will give you 20 Lumen, it seems to be pretty accurate.. it's maybe off 2 lumens at 1000mA.. and it'll likely have a similar error on the low side, but by in-large it does a really decent approximation of the curve on the spec sheet for the lux3.

besides being extremely useful if you don't have a light meter to estimate the brightness of your project to compare to other mods out there... it should be helpful for figuring out brightness settings with resistored or regulated mods that have multiple brightness settings.

here is the lumiled's chart:

and here is my caluculated chart based on that reference:

an interesting note is though my curve is lower on the high-end, the calculated value is actually higher.. i.e. at 1000mA, lumileds claims 72 lumen.. but their chart shows higher than that.. my curve is closer to what their text says, in any event it's a mighty fine approximation, hope it helps somebody else.. it was very nice for me since i don't have a light meter to get some approximation of light output during a runtime test.

-awr

 

[cy]

cool now we need to test this with an integrated sphere

 

[andrewwynn]

it would be awesome.. anybody that has one.. i would love if they would give me a few reference points and i can fine-tune the formula if necessary. (of course each emitter is a little different anyhow.. but it would be nice to have feedback)

 

[evan9162]

Did you take into account heat-induced dimming in your graphs?

I made a few of those several months ago, but took into account thermal modeling to give a true indication of output vs. current, and also provided junction temperature vs. current for various j-a thermal paths.

Here they are:

Output:

And Tj:

I also did the same for 1W and 5W luxeons.


Let's compare and see what the differences are.

 

[andrewwynn]

not sure i'm understanding that top graph.. says that if you cool it better you'll loose output?

I was looking at the relative output loss due to temperature as i was making that, but two main points... the typical use of a light and especially one with the type of heat sinking in my lights.. the Tj is probably about 20-30Deg above room temperature at worst case (since the initial heating doesn't typically even get the heat sink warm much less to a steady state).. and the idea for what i was intending isn't really an accurate plot of brightness long term, but to find out the brightness would be when it's first turned on at different current levels.

the long-term brightness factoring in the temperature is very interesting thoughts.. i have to figure out how to measure the Tj... I know that if i touch the heatsink directly opposite the emitter slug (and it's only 1/16th inch thick).. it never gets over about 40C in normal use... if i left it sit there like when i did the runtime test.. might have gotten up to 50 or 55. .it was quite warm.. but normally when i take off the battery and test for 'where is it hot'.. the heat is completely evenly distributed through the heat sink.. i can't even tell where it's hotter... it seems that as much heat comes from the battery as the emitter.

-awr

 

[evan9162]

[ QUOTE ]
andrewwynn said:
not sure i'm understanding that top graph.. says that if you cool it better you'll loose output?



[/ QUOTE ]

I guess I should have explained more. These come from here:
http://www.candlepowerforums.com/ubbthreads/showthreaded.php?Number=493239


Each plot represents a different thermal resistance from the junction to ambient. So the smaller thermal resistance (13C/W) represents the best thermal path, and thus, the greatest output at a given current. The other thing to consider is that the best (minimum) thermal resistance for an L3 is 13C/W - since the junction-slug resistance, which is a byproduct of the luxeon fabrication, is the minimum thermal resistance that is possible. The 13C/W plot represents the first few seconds of turning on the light, and a steady-state solution if you had a "perfect" heat sink held at 25C.

The vertical red line represents the current level at which the maximum junction temperature is exceeded.



[ QUOTE ]

I was looking at the relative output loss due to temperature as i was making that, but two main points... the typical use of a light and especially one with the type of heat sinking in my lights.. the Tj is probably about 20-30Deg above room temperature at worst case (since the initial heating doesn't typically even get the heat sink warm much less to a steady state).. and the idea for what i was intending isn't really an accurate plot of brightness long term, but to find out the brightness would be when it's first turned on at different current levels.


[/ QUOTE ]

Actually, regardless of what heat sink you have, the junction of an L3 will alwyas be 13C higher than ambient per watt of power dissipated in the luxeon. So, for 2.5W of power (typical of an L3 at 700mA), the junction, is at minimum, 33C above ambient. It's even more than that, since there's at least 1-2C/W of thermal resistance in addition to the junction-slug resistance (due to bonding the slug to the heat sink). Thus, the junction temperature is likely 40C above ambient at 700mA, and likely 60C above ambient at 1A.

The die of the Luxeon heats up in a matter of 20ms or so, so regardless of what heat sink, the 13C/W plot applies.


[ QUOTE ]

the long-term brightness factoring in the temperature is very interesting thoughts.. i have to figure out how to measure the Tj... I know that if i touch the heatsink directly opposite the emitter slug (and it's only 1/16th inch thick).. it never gets over about 40C in normal use... if i left it sit there like when i did the runtime test.. might have gotten up to 50 or 55. .it was quite warm.. but normally when i take off the battery and test for 'where is it hot'.. the heat is completely evenly distributed through the heat sink.. i can't even tell where it's hotter... it seems that as much heat comes from the battery as the emitter.


[/ QUOTE ]

There's no real way to directly measure Tj. The junction is actually sandwiched between two semiconductor layers. The best you could do for direct measurement is probing the top layer, but that would involve inserting a probe into the dome of the Luxeon, below the phosphor if it's a white. You can also guesstimate Tj differences by observing differences in Vf (pulsing vs. DC), but I've found that to not be so reliable, as I've attempted to control Tj and measure Vf differences, and delta-Vf vs. delta-Tj seems to vary on an individual LED basis.

However, for the measurements that you did - if we take the 40C to be the slug temperature, at 700mA, the junction temperature would be 73C; at 1A, 90C. So, with a 40C slug temperature, depending on drive current, you can expect between 15% and 25% less light due higher junction temperature.

 

[Doug S]

[ QUOTE ]
evan9162 said:
You can also guesstimate Tj differences by observing differences in Vf (pulsing vs. DC), but I've found that to not be so reliable, as I've attempted to control Tj and measure Vf differences, and delta-Vf vs. delta-Tj seems to vary on an individual LED basis.



[/ QUOTE ]
Thread hijack alert!!

Darin! I knew that if anyone else here had attempted to measure the Luxeon Tempco of Vf it would be you. The datasheets claim -2mV/C but I think this is bunk. I have tried repeatedly to measure the tempco and have come up with values ranging from -2.7mV/c to -3.8mV/C. I have no idea what probable error to assign to my techniques but what I considered my most rigorous testing yielded the -2.7mV/C. With that setup I didn't test multiple samples so I don't have a feel for the sample to sample variation. If you have or later generate some tempco data, please share it.

 

[cy]

what is the importance of junction temp?

 

[evan9162]

[ QUOTE ]
Doug S said:
[ QUOTE ]
evan9162 said:
You can also guesstimate Tj differences by observing differences in Vf (pulsing vs. DC), but I've found that to not be so reliable, as I've attempted to control Tj and measure Vf differences, and delta-Vf vs. delta-Tj seems to vary on an individual LED basis.



[/ QUOTE ]
Thread hijack alert!!

Darin! I knew that if anyone else here had attempted to measure the Luxeon Tempco of Vf it would be you. The datasheets claim -2mV/C but I think this is bunk. I have tried repeatedly to measure the tempco and have come up with values ranging from -2.7mV/c to -3.8mV/C. I have no idea what probable error to assign to my techniques but what I considered my most rigorous testing yielded the -2.7mV/C. With that setup I didn't test multiple samples so I don't have a feel for the sample to sample variation. If you have or later generate some tempco data, please share it.

[/ QUOTE ]
Doug,

I attempted a very thorough test to determine the tempco of Vf. I used 10 1W samples from the same tree (All Q2J).

I made 5 measurements (30, 130, 310, 620, 1080 mA) of each at 3 temperatures each (Tstar=-17C, Tstar=0C, Tstar=25C). I used a peltier to cool the star to the cooler temperatures, using watercooling on the hot side to achieve -17C, aircooling to achieve 0C, and using a plain heat sink for the 25C measurement. I measured the temp of the star board with my K-type thermocouple attached to my DMM to ensure consistant results (I recorded the actual temperature of the star board rather than assuming the same temp throughout).

So, I had 150 measurements I worked with (damn, I'm anal!). Basically, the tempcos were pretty darn consistent across the temp range for one sample, but they were all over the place between samples, from -2mv/C to -5mv/C. I was so fed up with the data I collected that I never bothered to record it beyond the piece of paper I used, which was promptly lost when I moved into my house a few months ago.

So, the tempco probably varies on a per sample basis, just like the Vf does, and just as much from my observations. One more source of random variation with these little beasties.

 

[andrewwynn]

importance to this thread.. not much ;-) the importance is not going over or it'll wreck the emitter.. the problem is.. (as the gurus above make clear).. it's very difficult to know what the Tj is since you can't measure it.

My point for the thread was just to have a general formula to estimate brightness compared to current on a lux3 using real values... i was trying to figure out roughly how flat the brightness curve was on a light i just built w/o having to use a light meter (which is about the next thing on my wish list).

I like the new info though.. tell me the details about C/W definition

-awr

 

[evan9162]

[ QUOTE ]
cy said:
what is the importance of junction temp?

[/ QUOTE ]

Managing the junction temperature is at the crux of keeping Luxeons cool. The junction is the layer of material within the LED die where photons are actually generated. The performance of the LED is dependent on the junction temperature.

As the junction temperature increases, the efficiency of photon generation (and therefore, brightness) decreases. Also, as the temperature increases, the voltage drop (Vf) of the LED decreases.

The junction temperature can be estimated by knowing how much power is dissipated in the LED, and knowing the thermal characteristics (thermal resistance) from the junction to ambient/heat sink/etc.

So, all of the effort in improving heat sinking has the ultimate goal of keeping the junction cool, as it is ultimatley responsible for producing light.

 

[cy]

Thanks for explaining the importance of the junction temp.

 

[gadget_lover]

[ QUOTE ]
andrewwynn said:
importance to this thread.. not much ;-) the importance is not going over or it'll wreck the emitter.. the problem is.. (as the gurus above make clear).. it's very difficult to know what the Tj is since you can't measure it.


[/ QUOTE ]

I think the gurus also said that you are probably experiencing up to a 25% drop
in light output if you are measuring 40C at your heatsink. This will have a direct impact on the curve. Where you show 80 lumens it's more likely to be closer to 60 due to elevated Tj. That should make a major difference in your chart.

Daniel

 

[andrewwynn]

yup.. i think i'll be going back and making a change.. however.. yours (and their) assumption is steady-state which is virtually never the case.. my heatsink temperature is normally 25C not 40C.. it only gets that hot during testing, and maybe one out of 20 times i use it.. that's the primary reason i opted not to factor that in.. my chart is a direct copy if the function from the spec sheet of the emitter, which presumably is at 25C and not factoring for temperature.

 

[NewBie]

Right, but the junction temperature is going to jump up immediately, say you put 1A into a typical Luxeon III with a Vf of 3.9V @ 1A. 3.9V * 1A = 3.9W.

Say you are using a Lux III with a 13 C/W rating. 13 * 3.9 = 50.7C Ambient is 25C, so 50.7C + 25C = 75.7C With the junction at 75.7C, you are going to see only 82% light output. LumiLEDs lumens is based on 25C junction temp.

As is the case with most flashlights, the body starts heating, and you see a further degredation of light output.

Add in a good 8-12% light drop for standard non-coated lens, and another 15-25% drop in light for a high quality vacuum sputtered Aluminum.

So in the case of a flashlight, 18% loss from heat initially, + 8%, +15% = 41% loss in best case, and 18% +12% + 25%= 55% loss. Then it starts dropping as flashlight body heats up.

You could go with UCL, and a very expensive laboratory reflector coating with enhancing oxide overcoats, so lens loss drops to 1-4%, and reflector loss drops to 2-7%.

For anyone that wants the datasheet, it is here:
http://www.lumileds.com/pdfs/DS45.PDF

 

[andrewwynn]

cool.. well minimighty is planned to have coated aluminum reflector and there is no way i'd make a light w/o a coated UCL lens...

In my test light... at 1A Vf is only 3.7V.. so 3.7W x 13 = 48.1C + 25C = 73.1.. so i start with a smidge higher output to start with.. and i have NO lens :-D so no loss there (of course.. that makes the reflector suck even if it was great to start with.. however that's all moot... the 'real' light will have UCL lens and a quality vacuum metallicized (sp) reflector..

It's very odd that lumileds would have a graph of luminous flux but not have the brightness loss factored in for the Tj.. the Temp of the slug should be the important figure in that design consideration.

thanks for the info.

 

[NewBie]

[ QUOTE ]
andrewwynn said:
cool.. well minimighty is planned to have coated aluminum reflector and there is no way i'd make a light w/o a coated UCL lens...

In my test light... at 1A Vf is only 3.7V.. so 3.7W x 13 = 48.1C + 25C = 73.1.. so i start with a smidge higher output to start with.. and i have NO lens :-D so no loss there (of course.. that makes the reflector suck even if it was great to start with.. however that's all moot... the 'real' light will have UCL lens and a quality vacuum metallicized (sp) reflector..

It's very odd that lumileds would have a graph of luminous flux but not have the brightness loss factored in for the Tj.. the Temp of the slug should be the important figure in that design consideration.

thanks for the info.

[/ QUOTE ]

Yeah, you'd figure, but remember marketing/sales has their spin, and they want to make the parts look as awesome as absolutely possible, and looking at datasheets, you have to look *very* carefully for things like this. Another thing is they like to hide "undocumented" features, so if something is excluded, there is usually a reason.

If you look at page 3 on http://www.lumileds.com/pdfs/DS45.PDF , you'll see they put Flux Characteristics at xxxx mA, actually spell out in the largest font on the page Junction Temperature, then Tj=25C (kinda like a double redundancy, alerting a person). They were trying to make it immediately obvious that something is up here that you are required to look into, then later on provide a chart so you can adjust as necessary. At least they give you the tools and information necessary to make a rather rough estimate on lumen output.

 

[evan9162]

Newbie's info is right on the money, but does paint a bleak picture as far as efficiency goes.

It seems that all of Lumileds info is at a Tj of 25C, which is basically impossible in normal conditions. The only way to achieve those results is with very short pulses, or supercooling. The datasheets show their products in the best possible light - and give enough info so one could figure out how they perform in normal conditions. While not really lying so to speak, the datasheets could definitely be more useful (and less misleading) by representing perforance under real-world conditions.

 

[cy]

[ QUOTE ]
NewBie said:[/b So in the case of a flashlight, 18% loss from heat initially, + 8%, +15% = 41% loss in best case, and 18% +12% + 25%= 55% loss. Then it starts dropping as flashlight body heats up.

You could go with UCL, and a very expensive laboratory reflector coating with enhancing oxide overcoats, so lens loss drops to 1-4%, and reflector loss drops to 2-7%.

[/ QUOTE ]

so less 18% for heat + 1% UCL + 15% lens = 34% loss in best case. UCL combined with PR 27mm and good heat sinking would fit this profile?

 

[Doug S]

I am apparently in the minority around here but I find no fault with the way Lumileds presents the information on their datasheets concerning output and the effects of drive current and temperature. There is *additional* information that I would *like* to see however. High on the list would be min and max limits in addition to the typical on that graph showing the output vs temperature curve. Also I would like to see min/max on that graph of output vs drive current. I have some experimental data showing that the degree of variation of output with current can vary considerably from sample to sample. I suspect, but do not have the experimental data to show, that the output vs temperature performance can vary from the graph shown on the datasheet. Maybe Darin can explore this in his spare time /ubbthreads/images/graemlins/poke2.gif
BTW, Darin, thanks for sharing you data on Vf tempco.