Oil cooled LED

[jtr1962]

EDIT: What was most surprising for me, was the idea that the probe gets hot because of the radiation it absorbs. I always thought that when I touch the thermocouple to the LED dome, what I measure is the actual dome temperature, not the massive amount of radiation that would be needed to rise the probe temperature...

You can vividly demonstrate this by putting a piece of black electrical tape over the dome of a Rebel or an XP-E/XP-G. The tape will start smoking almost immediately. On the other hand, do the same with a piece of white paper or clear tape, and nothing happens. This is telling me the dome doesn't get that hot in normal use. Rather, it is the emitted light energy being absorbed which makes some ( but not all ) objects placed near the dome get hot. And if you physically think about this it makes sense. Today's emitters can emit upwards of 100 lumens of light from a very tiny surface area. The efficacy of the emitted spectrum is around 330 lm/W, so 100 lumens equates to 100/330, or about 0.3 watts, of light energy. Think how a small component like a surface mount resistor gets when it dissipates 0.3 watts. And if you put a probe or black tape or your finger over the dome it will absorb roughly that much power. This is a relatively new, interesting phenomenon. A few years ago LED domes were much larger, and they emitted far less light, so the temperature rise experienced by objects in proximity to the dome was negligible.

But the fact remains that nearly 100% of the heat an LED produces is conducted to the thermal pad. Oil cooling of the dome will accomplish absolutely nothing.

 

[wapkil]

You can vividly demonstrate this by putting a piece of black electrical tape over the dome of a Rebel or an XP-E/XP-G. The tape will start smoking almost immediately. On the other hand, do the same with a piece of white paper or clear tape, and nothing happens. This is telling me the dome doesn't get that hot in normal use. Rather, it is the emitted light energy being absorbed which makes some ( but not all ) objects placed near the dome get hot. And if you physically think about this it makes sense. Today's emitters can emit upwards of 100 lumens of light from a very tiny surface area. The efficacy of the emitted spectrum is around 330 lm/W, so 100 lumens equates to 100/330, or about 0.3 watts, of light energy. Think how a small component like a surface mount resistor gets when it dissipates 0.3 watts. And if you put a probe or black tape or your finger over the dome it will absorb roughly that much power. This is a relatively new, interesting phenomenon. A few years ago LED domes were much larger, and they emitted far less light, so the temperature rise experienced by objects in proximity to the dome was negligible.

The light undoubtedly conducts energy but as I understand, we are talking about a small, fraction of a cubic millimeter size, thermocouple head. It also usually has a shining, metallic surface. It's hard for me to imagine how much radiation would have to be emitted around to significantly rise its temperature.

When I was measuring LED dome temperatures, the readings were starting to rise rapidly only after the thermocouple physically touched the dome. Moreover, after the light was turned off the measured temperature didn't drop rapidly as it would if the thermocouple was heated by radiation. I still suspect that it was a result of a simple heat conduction. It is probably sufficient to touch the dome after the light is turned off to see that it becomes quite hot

Nope. It will not make an appreciable difference in junction temps. Junction temperature is all that matters not the surface temp of the package.

But the fact remains that nearly 100% of the heat an LED produces is conducted to the thermal pad. Oil cooling of the dome will accomplish absolutely nothing.

I think I have to agree with you here. Even if I'm right that the dome is not exactly an insulator (AFAIR glass has thermal conductivity comparable to thermal pastes) it is definitely much worse than aluminum. What is worse, if I'm correct one should fill up the head with glass instead of water or oil to effectively transfer the heat in a relatively transparent medium and even with glass, in a properly designed flashlight the gain would probably be negligible.

 

[jtr1962]

The light undoubtedly conducts energy but as I understand, we are talking about a small, fraction of a cubic millimeter size, thermocouple head. It also usually has a shining, metallic surface. It's hard for me to imagine how much radiation would have to be emitted around to significantly rise its temperature.

Again, think about the physics of this. What really causes the temperature to rise here is the intensity of the light falling on the probe. Very close to the dome this can be two orders of magnitude higher than full sunlight! Granted, the probe may be small and not a perfect absorber. Nevertheless, it is still subject to the same light intensity at any given distance from the dome as black electrical tape. And because it has a small mass, it doesn't require as much energy to heat up. I already have noted this effect with my outdoor thermometer probe. It's white and therefore doesn't absorb much light. Nevertheless, in full sunlight it can be as much as 20°F higher than the shade temperature. Sunlight has a large component of infrared contributing to the heating, so this isn't a totally fair comparison. Nevertheless, I think visible light contributes at least 1/3 to the temperature rise, so if you want to correct for this you might say the probe rises up to 6°F due to visible light absorption. Close to the dome of a small power LED the visible light intensity can be two orders of magnitude greater than full sunlight. So that's potentally 100 times as much temperature rise. In reality it's less, because as an object heats up, it also starts convecting and radiating away some of the energy. The temperature it ultimately stabilizes at depends upon how well it absorbs energy versus how well it loses it as it heats up. Anything black and with low surface area (i.e. flat ) will get very hot in sunlight. Something like an aluminum heatsink with lower absorption and more surface area will experience far less temperature rise.

When I was measuring LED dome temperatures, the readings were starting to rise rapidly only after the thermocouple physically touched the dome. Moreover, after the light was turned off the measured temperature didn't drop rapidly as it would if the thermocouple was heated by radiation. I still suspect that it was a result of a simple heat conduction. It is probably sufficient to touch the dome after the light is turned off to see that it becomes quite hot.

That's the inverse square law. A mm from the dome the intensity might only be 1/3 as much as actually touching the dome. And the temperature didn't drop rapidly because of the thermal mass of the probe. Yes, it may be a small mass, but it also is a very small surface area from which to conduct away heat in order to drop in temperature. Besides that, there is very poor thermal conductivity between the probe and dome, even if they are touching. There would still be a lot of tiny air gaps and such without thermal paste. Because of this, any rapid temperature increase the probe experiences when the emitter is turned on is primarily due to it absorbing emitted light. I'll grant that the dome rises somewhat above ambient as it's not perfectly transparent. But certainly not too hot to touch. If anyone has a non-contact IR thermometer then perhaps they can help here. Measure the temperature of the dome immediately after the LED has bee turned off. That will negate any possibility of errors just in case the IR thermometer responds to visible light. It shouldn't, but then again I've never used one.

 

[wapkil]

Because of this, any rapid temperature increase the probe experiences when the emitter is turned on is primarily due to it absorbing emitted light. I'll grant that the dome rises somewhat above ambient as it's not perfectly transparent. But certainly not too hot to touch. If anyone has a non-contact IR thermometer then perhaps they can help here. Measure the temperature of the dome immediately after the LED has bee turned off. That will negate any possibility of errors just in case the IR thermometer responds to visible light. It shouldn't, but then again I've never used one.

Unfortunately I don't have an IR thermometer but I just tested a dome temperature of an MC-E LED driven with ~2.5A in a limited run L-mini II MC-E. After it ran for ~1 minute, I turned the light off, took a thermocouple showing the room temperature (22 deg. C) and touched it to the dome - it read ~50 deg. C. It took me around one second between turning off the light and touching the dome so the dome temperature was probably significantly higher immediately after the current was cut off...

EDIT: It also took some time before the thermocouple reading stopped rising. Obviously at the same time the dome temperature was falling, making the difference between the reading and the dome temperature when the light was on even higher.

 

[jtr1962]

Yes, the dome temperature probably fell a bit before the temperature stabilized, so perhaps it was 60° or 65°C when the LED was running. That actually sounds reasonable to me, and about in line with some of my other observations. For example, when I took an XR-E down to around -25°C with a thermoelectric module, the moisture condensing on the dome didn't freeze when the LED was running at 2+ amps, but did freeze when it was turned off. So the dome was probably 30°C or so higher in temperature than the heat sink. A smaller dome, being closer to the die, or a higher-powered LED with a large dome, like the MC-E you tested, would be higher in temperature still, perhaps as high as 65°-70°C at room temperature. This makes sense from a physical standpoint. Even if the dome only absorbs a few percent of the emitted energy, that's enough to heat it to those sorts of temperatures. But I wouldn't consider temperatures like this quite hot. I just ran a test on some P7s I had set up. I let them run a while, then shut them off and touched the dome. It's warm but not hot-I'd say maybe 50°-55°C. I can keep my finger on it indefinitely. However, if I keep my finger on the dome of one ( or just above it ) while it's turned on after about 20 seconds it's too hot to keep there. So it is indeed the radiation doing this heating, not the dome being too hot to touch. The instant I turn the LEDs off, my finger no longer feels hot. If the dome had been getting hot enough to burn my finger, then it would take a finite time for the burning sensation to go away as the dome gradually cooled.

What kind of temperatures do you get if you keep the probe on the dome and leave the LED on? I'd guess you'll be getting much higher temperatures.

 

[MarineBeams]

I have been testing an oil filled light for the past 10 months. 3 cree R2's in a IP68 sealed acrylic enclosure running 24/7 @500ma. The light was 5 feet underwater in my canal the whole time. The oil was 100% mineral oil and it completely surrounded the heatsink (a very small and thin aluminum flat-stock, no fins).

My theory: 1. Constant fresh water would cool the oil filled housing much better than air ever could, and the cool oil would keep the lights happy. 2. the oil makes the unit less likely to crush at extreme depths. 3. Water could not enter unless oil was expelled, making leaks less likely and a very small bit of water intrusion would not equal immediate death to the electronics.

The light just failed last week. I haven't cracked it open yet but can see a serious black buildup on the solder points, shorting the gap.

What I figure is that although the oil was 100% pure, impurities from the electronics and solder caused a slight electrolysis effect. Should have soaked them better with alcohol I guess.

I learned a lot from the experiment and have some idea's I may incorporate into another prototype.

 

[jtr1962]

I have been testing an oil filled light for the past 10 months. 3 cree R2's in a IP68 sealed acrylic enclosure running 24/7 @500ma. The light was 5 feet underwater in my canal the whole time. The oil was 100% mineral oil and it completely surrounded the heatsink (a very small and thin aluminum flat-stock, no fins).

My theory: 1. Constant fresh water would cool the oil filled housing much better than air ever could, and the cool oil would keep the lights happy. 2. the oil makes the unit less likely to crush at extreme depths. 3. Water could not enter unless oil was expelled, making leaks less likely and a very small bit of water intrusion would not equal immediate death to the electronics.

The light just failed last week. I haven't cracked it open yet but can see a serious black buildup on the solder points, shorting the gap.

What I figure is that although the oil was 100% pure, impurities from the electronics and solder caused a slight electrolysis effect. Should have soaked them better with alcohol I guess.

I learned a lot from the experiment and have some idea's I may incorporate into another prototype.

Try clear epoxy next time instead of oil. Most liquids eventually get through the packages of electronic parts, causing them to fail. The epoxy will harden, protecting everything against moisture and pressure.

 

[MarineBeams]

Try clear epoxy next time instead of oil. Most liquids eventually get through the packages of electronic parts, causing them to fail. The epoxy will harden, protecting everything against moisture and pressure.

That was my first choice, but finding a non-yellowing epoxy with no air bubbles and good (great) transparency was a stumbling block. If you know where to get some I will give it a try!

 

[Daekar]

I was considering using mineral oil to create a non-conventional heatsink solution for modding a very old sheet-metal incan. While the spirit of the light moved me to lean in the direction of hotwire, I don't see why oil couldn't have helped in the absence of other heatsinking solutions, particularly when there was so much surface area to conduct to. Oil solutions seem to work extremely well for PC overclockers, too.... just immerse everything but the hard drive in mineral oil, hook up a remote radiator and pump, and they're good to go.

It is good to hear (although I'm sorry about its untimely demise) about possible stumbling-blocks revealed by your light, MarineBeams. I wonder if the combination of more thorough alcohol cleaning combined with a thin epoxy coating over most of the solder joints would prolong the lifespan of the system?

I don't think an oil-light is necessarily a good solution for all light designs and purposes, but in non-traditional designs it seems like something worth considering.

 

[wapkil]

Yes, the dome temperature probably fell a bit before the temperature stabilized, so perhaps it was 60° or 65°C when the LED was running. That actually sounds reasonable to me, and about in line with some of my other observations. For example, when I took an XR-E down to around -25°C with a thermoelectric module, the moisture condensing on the dome didn't freeze when the LED was running at 2+ amps, but did freeze when it was turned off. So the dome was probably 30°C or so higher in temperature than the heat sink. A smaller dome, being closer to the die, or a higher-powered LED with a large dome, like the MC-E you tested, would be higher in temperature still, perhaps as high as 65°-70°C at room temperature. This makes sense from a physical standpoint. Even if the dome only absorbs a few percent of the emitted energy, that's enough to heat it to those sorts of temperatures. But I wouldn't consider temperatures like this quite hot. I just ran a test on some P7s I had set up. I let them run a while, then shut them off and touched the dome. It's warm but not hot-I'd say maybe 50°-55°C. I can keep my finger on it indefinitely. However, if I keep my finger on the dome of one ( or just above it ) while it's turned on after about 20 seconds it's too hot to keep there. So it is indeed the radiation doing this heating, not the dome being too hot to touch. The instant I turn the LEDs off, my finger no longer feels hot. If the dome had been getting hot enough to burn my finger, then it would take a finite time for the burning sensation to go away as the dome gradually cooled.

What kind of temperatures do you get if you keep the probe on the dome and leave the LED on? I'd guess you'll be getting much higher temperatures.

The temperatures I would get if I kept the probe on the dome would be something around 20-30 deg. C higher. I agree that the radiation can heat up the dome - all the radiation has to travel through it and probably a few percent is converted to heat. What I'm still not sure about is how much the radiation alone can rise the temperature of the thermocouple.

It's an interesting topic for me because the LED dome is frequently the only place on the LED that can be easily measured in an assembled flashlight. The question is, whether it can be used to estimate the junction temperature.

When I was measuring the dome temperature in an MC-E LED powered with ~10W, the comparison to the junction temperature estimated from the theoretical junction temp./brightness relation suggested that the difference between the junction and the dome is only ~20 deg. C. It was surprising for me, because I believe there shouldn't be such a low resistance thermal path between the junction and the dome.

Maybe the temperature rise due to the radiation absorbed by the dome can be the answer? It is also possible (although seemed unlikely to me) that the thermocouple doesn't show the actual dome temperature if it is indeed additionally heated by the radiation. I haven't investigated it further and I don't know the answers.

 

[jtr1962]

The temperatures I would get if I kept the probe on the dome would be something around 20-30 deg. C higher. I agree that the radiation can heat up the dome - all the radiation has to travel through it and probably a few percent is converted to heat. What I'm still not sure about is how much the radiation alone can rise the temperature of the thermocouple.

Radiation at the intensity existing at the dome can easily raise the temperature of the thermocouple 20 or 30 degrees C. Like I said, it's up to two orders of magnitude higher than full sunlight. We all know how being in the sun can raise the temperature of things above ambient. Same mechanism applies here.

It's an interesting topic for me because the LED dome is frequently the only place on the LED that can be easily measured in an assembled flashlight. The question is, whether it can be used to estimate the junction temperature.

Interesting question, but I can tell you right now dome temperature and junction temperature are not really related.

When I was measuring the dome temperature in an MC-E LED powered with ~10W, the comparison to the junction temperature estimated from the theoretical junction temp./brightness relation suggested that the difference between the junction and the dome is only ~20 deg. C. It was surprising for me, because I believe there shouldn't be such a low resistance thermal path between the junction and the dome.

There isn't a low resistance thermal path between the junction and dome. The primary mechanism through which the dome heats up is absorption of some of the LED's emitted light. If the dome were perfectly transparent then it would be more or less at ambient temperature, perhaps a few degrees above. The dome is thermally coupled to the LED case, but it's relatively poor coupling because the dome material is a poor thermal conductor relative to the other materials in the LED case ( usually alumina ceramic and copper pads ). It's sort of like if you put a piece of styrofoam insulation on a hot metal plate. You can easily touch the styrofoam even if the plate is 100°C. The LED dome isn't as insulative as styrofoam, but the same general idea applies here. It wouldn't be much warmer than ambient if it didn't absorb some light ( and it absorbs only a few percent at most ).

Maybe the temperature rise due to the radiation absorbed by the dome can be the answer?

That is definitely the answer. It would be nice if there were some external way to measure junction temperature but unfortunately there isn't.

 

[wapkil]

Thank you for the explanations. I'd have to check but I think that with 10W power even a few percent of the radiation absorbed may translate to a few hundred milliwats heating the dome.

Technically there has to be some relation between the junction and the dome temperature but it may be much too complicated to be useful. For example if it works as you described, in the same setup a lower junction temperature could lead to a higher emission and a higher dome temperature

On the other hand, if in a particular setup this relation could be established, it could probably also be used to estimate the junction temperature, e.g. in my L-Mini measurements it seemed pretty consistent with theoretical temperature/brightness values. I'm not sure though if what I did there was completely correct - I'll have to add a link to this discussion in that thread.

BTW, do you know any better way of estimating the junction temperature in an assembled flashlight than measuring the changes in brightness?

 

[jtr1962]

That was my first choice, but finding a non-yellowing epoxy with no air bubbles and good (great) transparency was a stumbling block. If you know where to get some I will give it a try!

I haven't tried it yet, but this stuff looks pretty good.

 

[Freeze_XJ]

Then still the heat transport would be low. Oil absorbs quite some heat, but sucks at transmitting it. To get effective cooling, you would want to circulate the stuff (completely impossible with epoxies), which isn't funny if we're talking oil. Just a little bit of flow tremendously increases cooling capabilities, but non-moving oil (or water, for that matter) is almost as horrible as air.
Just for your referencing :
Conductivity of air : 0.024 (W/mK, Watt per meter per kelvin)
Aluminium : 250
Oil (olive) : 0.17
Oil (machine stuff) : 0.15
Glass : 1.05
source
So if you really want to have static oil : use glass instead.
Or even better : make your entire substrate out of metal, and call it the anode Isolate a small patch to the backside as kathode, and connect your anode to the outside of the light. Best conductivity you can possibly get.

 

[JFD140]

there is some secret trick to using the clear max clr... i returned the container i ordered because its damn near impossible to get all the air bubbles out of it. Which leaves you with a bubbly covering.
The epoxy also throws off the optics of the LEDs dramatically reducing brightness.

 

[Illum]

Then still the heat transport would be low. Oil absorbs quite some heat, but sucks at transmitting it. To get effective cooling, you would want to circulate the stuff (completely impossible with epoxies), which isn't funny if we're talking oil. Just a little bit of flow tremendously increases cooling capabilities, but non-moving oil (or water, for that matter) is almost as horrible as air.

Considering there are miniature PC fans that are IP55 and IP21 rated it wouldn't necessarily be impossible to mount a fan in the "oil cell" and flood it with oil

 

[kidsonp]

MarineBeams - have you got any further with your experiments of running an LED in oil? It' s been a while but i'm still interested.

Ta

 

[Norm]

MarineBeams - have you got any further with your experiments of running an LED in oil? It' s been a while but i'm still interested.

Ta

:welcome:

Unfortunately MarineBeams doesn't appear to have visited CPF in almost three years.

Norm

 

[blasterman]

There was a liquid parafin cooled LED prototype being talked about awhile back.

Same problem - once you heat the liquid up, you then have to remove the heat from the liquid.

 

[wquiles]

Same problem - once you heat the liquid up, you then have to remove the heat from the liquid.

+1

It is "always" the same problem. The more power in the LED, the more heat you have to remove (somehow).

The process is pretty much also always the same:
- LED heats up
- the LED efficiency drops, the output of the LED starts getting lower,
- the vf lowers and the current goes up *(unless you have a current regulated power source)
- thermal energy travels directly or via MCPCB to heatsink
- heatsink heats up
- thermal energy travels from heatsink to body or head of light

-> at this point only you now have just two ways to transfer heat outside of the body:
- the poor transfer from body to the external air
- to the blood of the person holding the light

Either way, as the LED keeps running, the system (flashlight by itself in air, or flashlight in the users' hand) will reach a steady state temperature.

The higher the power being fed to the LED and the smaller the host, the quicker the person holding it will find the host too hot to handle comfortably for long periods of time. The larger the host, the longer it takes for this to happen, plus there is more surface area to help dissipate some of the heat (but again, thermal transfer to air is terrible as air is a good thermal insulator).

NOTE: The thermal resistance at the various junctions delays the transfer of heat, but the thermal energy still needs to travel and be dissipated (somehow). Yes, Copper is more efficient, but you are just moving it quicker to the body - Copper is not somehow "eliminating" the heat - it still needs to be dissipated (somehow). So it maters not if you have Al, Copper, fluid, etc.. - the amount of energy that needs to be moved is "still" the same. You are just delaying the steady-state system temperature.

Will