Thermal regulation results of the Arc4

Today, I got a spare moment and did more tests on the Arc4 thermal regulation. My goal is to verify that these units are always operating the LED at a safe level. As you know, temperature is the most damaging factor to LEDs. It affects the thermal maintenance, efficiency, tint and reliability. These are all factors dear to Arc.

So I took a Arc4x which drives the LED harder than a 4+ and performed time/temp measurements on the housing and the LED (the lens was removed so the IR pryometer has a fairly direct shot to the LED die). For all tests, the flashlight was placed on it's side on a wood desk. The pyrometer was aimed directly at the LED and moved around slightly to find the highest reading. Understand that using an IR pyrometer is the most practical method of measuring relative LED temperature once the LED is intalled in a flashlight. Figure that between the die and the IR image out the front there may be as much as a 15 degree gradient worse case.

#0642 4x 3/9/04
----------------
00 minutes - 26c LED - 26c housing - L1
02 minutes - 45c LED - 35c housing - L1
05 minutes - 52c LED - 42c housing - L1
07 minutes - 57c LED - 45c housing - L1
10 minutes - 58c LED - 49c housing - L2
12 minutes - 60c LED - 50c housing - L2
15 minutes - 55c LED - 49c housing - L2
17 minutes - 56c LED - 49c housing - L2
20 minutes - 57c LED - 49c housing - L2

Note that housing never went above 51c because the housing is calibrated for a max of 55c. You can also see that the thermal gradient between the LED and the housing is only about 10c, which is pretty good.

The manufacurer of the LED (Lumileds) states in their DS25 doc an average 70% lumen maintenance over 50k hours if the die temperature is kept at or below 90c. Obviously the lower the temp is kept, the longer the LED will last.

Peter

Thanks for the information. It is nice to confirm that I will probably be long gone before the LED in my Arc4's.

Peter, you and I once had a long exchange about the validity of this testing methodology for measuring the LED die or slug temperature. My position then, as now, is that this method is not valid. The radient emissions of the LED only occupy a small portion of the instrument's field of view. Whatever the instrument's viewing angle [aspect ratio], it increases a very close range. Same principle applies as when you view something held 1" from your eyeball, it severely defocuses. My $0.02.

Hello Doug,

Are you saying that the trend that Peter is showing is invalid or that the actual numbers may be not be correct?

Tom

[ QUOTE ]
SilverFox said:
Hello Doug,

Are you saying that the trend that Peter is showing is invalid or that the actual numbers may be not be correct?

Tom

[/ QUOTE ]
Actual numbers are suspect. Body temps should be OK.

Doug, I appreciate your input and you are welcome to produce evidence to the contrary. You said, "actual numbers are suspect". Since I gave only relative numbers then I don't see what new information you provide.

As it is, I have estimated a larger gradient between the top of the LED package and the die (15c) than I have measured between the package and the housing (10c).

Also, the reflector collimates the IR into the sensor quite well.

I maintain that the Arc4 is keeping the LED below 90c at all times and I am open to anyone proving otherwise.

In fact, if anyone can scientifically prove that an unaltered 4+ (First) is driving the LED die over even 80c, I will give them an LSH-P for free.

/ubbthreads/images/graemlins/smile.gif

Peter

Hello Peter,
off-topic, but you're saying that you took an arc4-x for the test: how many do you have in stock ? Perhaps it's time to sell one of these .

[ QUOTE ]
Francois1 said:
Hello Peter,
off-topic, but you're saying that you took an arc4-x for the test: how many do you have in stock ? Perhaps it's time to sell one of these .

[/ QUOTE ]

I was thinking the same thing. /ubbthreads/images/graemlins/grin.gif $195 right? /ubbthreads/images/graemlins/wink.gif

[ QUOTE ]
Gransee said:
...In fact, if anyone can scientifically prove that an unaltered 4+ is driving the LED die over even 80c, I will give them an LSH-P for free.

/ubbthreads/images/graemlins/smile.gif

Peter

[/ QUOTE ]
Is that with or without stripes? /ubbthreads/images/graemlins/smirk.gif

How about we attach a K-type thermocouple directly to the side of the luxeon slug (removing a section of the black plastic first)?

The pyrometer method is doubtful for many reasons, and the accuracy would be *highly* suspect.

As I recall, Gransee said the microprocessor temperature sensor that is on the microprocessor board kept the circuit board below 40C?

Whats the ambient air temperature?

Whats the flashlight body temperature after 5, 10 minutes?
In free air?

Whats the slug temp at the same time?

If the luxeon is being driven at 2W, at 15C/W, the LED die should be 30C above the slug temperature, and the slug would be higher than the flashlight body temperature.


Okay, take your 42 C housing temperature. With 2W to the LED, the die temperature should be at a minimum of 72C, and most likely, it is even higher.

very highly suspect, if not impossible.
==> 07 minutes - 57c LED - 45c housing - L1

Whether or not his equipment and method are completely accurate is up for endless speculation. However, I do not believe that the readings are that far off as to be over design spec for the led. Like he said, free lsh if led is proved over even 80c.

In order to know the junction temperature, you must know the drive current of the LED, plus the Vf of the led at that drive current.

Then, you must know the temperature of the slug. If you know the temperature differential between the body and the slug, then this will work too.

So, to calculate junction temperature, you would do the following:

Tj = T(body) + delta-T (body-slug) + (Vf * If * 15)

Random numbers:

If T(body) is 50C, and delta-T (body-slug) is about 3C, Vf is 3.6V, If is 600mA, then the junction temperature is:

50 + 3 + (3.6 * 0.6 * 15) = 85.4C

Higher Vf and higher If will cause the junction temperature to increase. Lower Vf and lower If will cause the junction temperature to decrease.

Even if you eliminate the slug-body thermal resistance, the junction temperatuer will still be above 80C in the above example.

Peter, if any of the ARC4s use a drive current of 600mA or more, contact Evan9162 for his shipping address /ubbthreads/images/graemlins/tongue.gif

/ubbthreads/images/graemlins/popcorn.gif

"close, but no LSH-P" /ubbthreads/images/graemlins/smile.gif

See Henry's reply on this...

"The calculations would be correct if all the assumptions were correct. However, the assumptions are not quite correct. Allow me to explain how this all works.

First, the thermal regulation - this is how it works today and, as always, is subject to change. The light thermometer is calibrated for 55 degrees C. Starting at about 47 degrees C, the light starts restricting power output by changing to the next lower setting. There is about 2 degrees C per setting restriction. So this means that setting 1 only operates up to 47 degrees C, setting 2 up 49, setting 3 up to 51, setting 4 up to 53, setting 5 up to 55. Setting 6 and above have no restrictions because there is insufficient power to cause damage - if the light is that hot, there is an external source of heat.

Second, there is a time factor (hence the term thermal -impedance-). At T=0, everything is at the same temperature so the delta-T from the junction (let's assume it is suddenly 35 degrees C hotter than the slug and case) is at a maximum and thus provides the best cooling, quickly dropping the junction temperature. At T a few minutes later, the temperature of the slug and interior of the heat sink have gone up and there is now a smooth temperature gradient from the junction to the surface and thus the junction temperature will coming to its maximum. If we can hold the case at a stable temperature, the gradient will eventually stabilize and so will the junction temperature.

So how hot will the junction actually get? Our first assumption is power to the LED - around 1.8 - 2.2 watts seems to be typical. This tells us that the thermal gradient across the slug will be 27 to 33 degrees C using data published by Lumileds. We also need to add something for the slug to heatsink interface and the temperature change across the heatsink - 3 degrees C was used below and that seems like a fair number. So if we add it all up, we get a junction temperature of 77 to 83 degrees C. But we are not done yet.

The temperature sensor is very close to the slug - closer to the slug than to the exterior. And in general, you should assume the temperature is rising. Therefore the temperature trip point will be lower than the case temperature by 1 or 2 degrees C. So the final junction temperature at trip will be in the 75 to 82 degrees C range. And of course, you have to allow for production tolerances and measurement errors. This is a worst case scenario and under real conditions, it is also short lived - someone sets the light down, it runs up to thermal limit and trips down to the next lower setting.

Under typical usage, the case temperature will remain below 40 degrees C and thus the junction temperature will remain below 80 degrees C.

Henry."

Henry showed it was possible to get a junction temp above 80c but said it was unlikely in normal use. And the example I measured was an 4x while I asked for proof with the 4+, which drives the LED at a lower power.

There is a difference between saying something is possible (theory) and actually proving a real world instance (fact). I would at least need photos, etc. For the LSH-P I was willing to spend, I was hoping someone could provide real world test results. But I have yet to see those. I saw the expense of the LSH-P as an investment in proof.

When a theory provides enough evidence that the phenomena is likely to not occur then it is hardly a scientific fact until independantly verifiable experiments can provide proof one way or the other.

Even the theoretical results show that the part is more unlikely to exceed the maximum temperature put forth by the LED manufacturer. I feel that this, "more unlikely" is sufficient for me to invest an LSH-P. /ubbthreads/images/graemlins/smile.gif

The Arc4 may be the only light to provide this level of brightness in this size package while keeping the LED as cool as is does.

Peter

Well, I certianly never had the expectation that my short response would garner an LSH /ubbthreads/images/graemlins/tongue.gif So no worries about me trying to argue my way to a free light /ubbthreads/images/graemlins/wink.gif

I'm convinced that it's quite easy (and likely does happen) that the junction temperature will go above 80C, but only momentarily, as it will drop again right when power levels are cut. I didn't mean for my equations to be taken as a steady-state model; only a worst case, right before throttling model.

What is the maximum current that an Arc4+ will ever put through the emitter? (Or would that be telling too much? /ubbthreads/images/graemlins/wink.gif )

I think I asked about this ( the ambient temperature question) before, but never got a firm reply.

[ QUOTE ]
Peter quoted Henry as saying

Under typical usage, the case temperature will remain below 40 degrees C and thus the junction temperature will remain below 80 degrees C.

[/ QUOTE ]

If a 4+ is used when the ambient temperature is high, let's say 44 C to 50 C, would the higher settings be essentially unavailable due to the unit's thermal protection? This might occur during daytime in enclosed places (especially in desert regions or when the unit is left in a vehicle). At 47C ( 116.6 F) it would seem that the temp circuit would almost immediately restrict the power flowing to the LED. I realize that this is no big deal, unless one plans on using the 4+ immediately after removing it from a vehicle (like in the glove box) or when working in an enclosed, high temperature setting (inspecting truck cargo, working in attics, etc.). Also, under extreme conditions like these, might it be possible for the LED to exceed spec briefly, before temperature regulation takes control?
I, for one, would much prefer some temperature control (with its negative impact on output level), than have to worry about damaging the LED under extreme conditions. But I am curious as to Peter (or Henry's) opinion on this.

[ QUOTE ]
Gransee said:
The manufacurer of the LED (Lumileds) states in their DS25 doc an average 70% lumen maintenance over 50k hours if the die temperature is kept at or below 90c. Obviously the lower the temp is kept, the longer the LED will last.

Peter

[/ QUOTE ]

This seems quite relevant to this new discussion:

Interesting, actual independent tests on the Luxeon, by the Lighting Research Center show a 10% drop in light output with a 45 degrees C die temperature (30 degree C slug(case) temp)after only 9,000 hours.

(see http://www.lumileds.com/pdfs/protected/AB07.PDF Fig 3 write-up)

Then LumiLEDs shows 10% drop in light output at 2,000 hours with a die temp of 105C. (see http://www.lumileds.com/pdfs/protected/AB07.PDF Fig 5). So heat is a definitely a major player.

So the 10% point is somewhere between these two.

Additionally, Though I do wonder if *Temperature* is the only factor, especially with the 500 hour "lifetime" of the 5W Luxeon. (see http://www.lumileds.com/pdfs/protected/DS40.PDF first page)


It has been nearly two years since the Lighting Research Center report came out, and no further updates have been forthcomming (odd).

Another worthwhile figure to look at is the Luxeon III emitter datasheet and Figure 5. It is found here, http://www.lumileds.com/pdfs/protected/DS45.PDF Its the current derating curve based on ambient temperature and thermal resistance from junction(die) to ambient. Why it is useful, is that it includes information at *700mA*.

Gransee you left out a key part...

"Average Lumen Maintenance Characteristics Lifetime for solid-state lighting devices (LEDs) is typically defined in terms of lumen maintenance—the percentage of initial light output remaining after a specified period of time. Lumileds projects that Luxeon products will deliver on average 70% lumen maintenance at 50,000 hours of operation. This
performance is based on independent test data, Lumileds historical data from tests run on similar material systems, and internal Luxeon reliability testing. This projection is based on constant current 350mA operation with junction temperature maintained at or below 90°C. Observation of design limits included in this data sheet is required in order to achieve this projected lumen maintenance. "
http://www.lumileds.com/pdfs/protected/DS25.PDF page 11

If the unit is heat regulated couldn't a software plug-in be designed for the flashlight to out put the highest internal temperature recorded by the flashlight firmware in CW? It must measure the temperature. So that figure should be available some where in a register in the memory of the flashlight. You know like telemetry.
Otherwise how tough would it be to build a testbed ARC 4+ with heat sensors in the heat sink and at the base of the slug. Just need to one as a typical unit you could then test the light with different body's as well to determine how well heat is managed.
Just a thought.
Yaesumofo

Great idea yaesomofo.