Modified star for increased thermal conductance

I have just started building custom lights, and the heat generated by an XML at 2.8 amps in an EDC-size flashlight has led me to the conclusion that I need to put more thought into thermal management than just trying for a thin bondline when gluing a star to the heatsink with arctic alumina.

My first thought was to drill a hole in the center of the thermal pad on a star (one of the 8mm XML bare stars made by Datiled) that would allow the thermal pad of the emitter to be soldered directly to a copper heatsink with a small pedestal the size of the hole, so that's what I did. I drilled a 7/64" diameter through-hole centered on the thermal pad, and then tinned the star, and mounted an XML (first time reflow soldering, pucker factor was a little high). I used a lab hotplate and a thermocouple to manage the heat, and it seemed to work out, as the emitter lit up afterward.

Next step is to solder the star to a copper heatsink by way of the hole, which is about 0.0625" deep and now partially filled with solder from emitter mounting. I don't have any way to measure any increase in performance, but I wanted to put this out there to see if anyone had already done it, and if they found it was worth it for the increase in performance/reliability/comfort. I will have pictures up tomorrow; I'm having trouble getting a decent shot of such a small subject with out point and shoot.

One question I have for the resident thermal experts is whether I want to minimize the thickness of the solder layer between the emitter pad and the pedestal on the copper heatsink, which will be the only attachment point holding the star to the heatsink, or if I can forgo the pedestal and keep the heatsink flat. I will be lapping the star base flat to match a shallow pocket in the heatsink, and may or may not put thermal grease between the star and heatsink (surrounding the solder "via"). Sorry if this makes no sense, I will have pictures up soon.

Line breaks are always awesome

But other than that, its a solid concept and direct soldering has definitely been done before in various ways. Some MCPCB makers even offer to put a hole through the insulating layers and plate it full of copper for non electrical pads so you can have the LED or power device soldered directly to your substrate with no insulating layer.

The amount of power put into an XM-L, and the low (thermal) mass of it means that in seconds the thermal resistance starts to lower the output. I saw a 2k lux improvement by switching from a cutter to a KD mcpcb with all the other test setup the same, and even repeated the experiment and had the same results. Directly soldering the thermal pad to a copper slug, without it having to conduct though a (hopefully) very thin insulating layer on a MCPCB would definitely be an improvement. Would you be able to measure the difference in output with a lux meter? Most likely. Would you notice it with your eye... questionable, and a point of debate at times.

Soldering directly to the heatsink in my opinion is a great idea! I must say thought that if the heat sink is small or it does not transfer heat away from the LED to the flashlight body efficiently, then the difference in thermal transfer would not be very significant.

Thermal conductivity of solder is low compared to copper (385) (http://www.electronics-cooling.com/2006/08/thermal-conductivity-of-solders/) so I'd assume it's best to keep the layer of solder as thin as possible.

Thanks for the replies, and yeah, when I write at 1 am sometimes it results in sort of a stream-of-consciousness thing. I went back and broke it up a little.

Norcimbus said:

... the heat generated by an XML at 2.8 amps in an EDC-size flashlight ...

Click to expand...

All that we can do in an LED light is move this energy through the various thermal barriers and reach the host, then reach a steady state or temperature equilibrium.

At a high level these barriers are:
- LED die to LED mounting base (typically meant to be soldered) thermal barrier
- mounting base LED to LED star thermal barrier (optional, but makes soldering the LED easier)
- mounting base (or star) to heatsink thermal barrier
- thermal barrier between the heatsink and the body/host of the light
- thermal barrier between the host and its surroundings, which could be:
1) Air - light sitting on a table, for example. Worst possible scenario as still air is a great thermal insulator
2) human hand - the hand is holding the light. The blood circulating on the hand cools down the light, up to a point where the light becomes too uncomfortable to hold. This is directly related to the thermal mass of the light. A smaller light (EDC) with 10 watts will get hotter quicker (and therefore get more uncomfortable quicker) than a large light (Mag 3D) also with 10 watts.
3) Water, as in a diving light. Best case scenario, since the thermal capacity of the water is basically infinite. The water can remove heat non-stop, so the steady state temperature of the LED under watter will be lowest possible, compared to cases 2 and 1 above.

Note that the end result, or steady temperature, will depend greatly on the ambient temperature and the heat capacity of the medium to which the LED is being used/exposed.

The size of the host affects how quickly the host will get hot, because of thermal mass. A larger light has a larger thermal mass, so it will take longer for the thermal energy to spread through a larger host. The larger host also has a correspondingly larger surface area, so a larger host has a better chance to exchange heat to its outer surface (air, blood, water).

Why the long speech? I am just trying to set the right expectations. 10 watts is 10 watts. A small EDC light with 10 watts will get hot quicker than a larger host with 10 watts. Increasing the thermal efficiency of the LED to heatsink only increases how quickly and efficiently the heat will travel to the heatsink and the host - it does "nothing" to make the EDC host more comfortable.

The only thing that can make a small EDC host more comfortable is to reduce the amount of energy being fed to the LED. Since about 80% of the energy is transformed to heat by the LED, the less hard you drive the LED, the lower the steady temperature will be, and the more comfortable holding the light will be.

Will

I'm wondering what you mean by 'building custom lights'. Do you mean you have a lathe and you're making custom heat sinks and or hosts or do you mean that you are modifying existing lights and only have some hand tools?

As for the thermal path, You drilled out the star and now the LED connects directly to the copper HS that has a little part that comes up through the star? Ya i think that would help some. If you have a significantly long enough thermal path through solder then it is counter productive since solder has a much lower thermal conductance that copper or Al. (So I am not 100% if you drilled a hole and filled it with solder or copper. Copper is good, solder is bad.)

I'd think about it this way:
Copper has a very high thermal conductivity (~400) so you want the thermal path to the host to be mostly copper.
Aluminum is decent (at around 230. less, maybe 130 depending on if it is an alloy) so having some of it is ok. People make whole heat sinks out of it so it really isn't too bad.
Solder isn't great (~50) but used as a connector in small amounts it is ok and can be needed for electrical connections.
Thermal glue/epoxy is pretty low (5 ish we will say). So you want only thin amounts and only if needed.
Air sucks (0.025) so avoid it when you can by using thermal glue (or a better fitting set of parts).

So the very best thing should be the LED base mounted directly onto a copper HS (with no star) (a copper star would be pretty good too, I wonder if anyone makes those). If you can figure out how to wire that up properly it is your best bet.

As wquiles mentioned, you still have a small piece of metal (the flashlight itself) that is putting out nearly 10 watts of heat, it will get hot if not cooled down by a hand or wind or something.

So I was thinking about wquiles's post a bit, and how he only talked about the barriers, ie. the interfaces from one object to another. So I got to thinking, what if my thermal epoxy had a conductivity of 4 and was 0.1mm think. Well that would provide a similar amount of resistance to 10 mm of copper. It makes me think about how a good quality epoxy can make a difference, and about how you want as think a layer of it as possible.

So long story short I agree that the barriers between the materials are very important to take into consideration too, so if you can avoid thermal paste it can be helpful.

The Thermo conductivity of Solder is 3-4 times worse than aluminum,
by grinding a hole in Aluminum & fill it with solder, I think the mod is making the juntion thermoconductivity worse.

However, 2.8A isn't a lot, probably won't see a diffrence at that current.

Okay, so no increase in comfort for EDC-sized lights. At what size then might a modification like this start to make a difference in performance or longevity? To clarify, I am thinking of making a raised post on the heatsink to fit into the hole drilled in the mcpcb, so the solder layer should be relatively thin, lets say ~.020"

Norcimbus said:

Okay, so no increase in comfort for EDC-sized lights. At what size then might a modification like this start to make a difference in performance or longevity? To clarify, I am thinking of making a raised post on the heatsink to fit into the hole drilled in the mcpcb, so the solder layer should be relatively thin, lets say ~.020"

Click to expand...

There is no size that will make a light comfortable - it is all about the amount of watts being dissipated. Nothing you can do in the mounting system, LED star, heatsink, etc., will change the amount of watts - all of those optimizations just change how quickly and efficiently the heat will reach the body.

A larger host will take longer to warm up, but it will warm up eventually. The larger size host also has more surface area to help radiate heat, but at the end of the way, it is all about the wattage, and how efficiently this heat can be removed away from the host.

If the main way for the heat to be removed is your hand (the blood in your body taking away heat from the host while you are holding your host), then it is a matter of how hot can the host get before "you" decide the host is too hot. In a medium-large host, like a 1D or 2D Mag host, "I" feel that about 10-15 watts is about the max. I can stand after the host is fully warmed up. For a small host, like an "E" size host, even 5-7 watts will get uncomfortable after 5-10 minutes, which is why the better sellers of small/EDC lights always advice to use the high more sparingly.

Edit: In this thread, post #31, I go through math to find out what size heatsink you need to achieve a target steady state temperature, based on the wattage of the LED(s):
http://www.candlepowerforums.com/vb/showthread.php?321319-CNC-Milling-Services/page2

About 49°C (about 120F) "should" be the max. steady temp to prevent getting burned, since in "about" 1-5 minutes you could get low degree burns:
http://www.armstronginternational.com/files/products/valves/pdfs/ay-699.pdf

Will

I will look into that calculation thanks Will. I think your explanations have put the comfort issue to rest, for me anyway.

Other than comfort, do you have an opinion about the value of this mod for increasing output or longevity of the emitter? Or are mcpcbs really the best option for most lights? I'm beating this issue to death because there seems to be a lot of interest in using copper heatsinks over aluminum heatsinks, as well as surface mounting the emitter, all with the purpose of maximizing performance.

Norcimbus said:

Other than comfort, do you have an opinion about the value of this mod for increasing output or longevity of the emitter? Or are mcpcbs really the best option for most lights? I'm beating this issue to death because there seems to be a lot of interest in using copper heatsinks over aluminum heatsinks, as well as surface mounting the emitter, all with the purpose of maximizing performance.

Click to expand...

The mods to increase the efficiency of the heat transfer, specially right at the LED, allow the LED to remain a little bit cooler, which allows the LED to have a higher lumen output. If you look at mods by saabluster you will find how effective he is in getting the max. output of the LED, by optimizing the heat transfer to the heatsink and the body. Very impressive in my honest opinion.

Will

ma_sha1 said:

The Thermo conductivity of Solder is 3-4 times worse than aluminum, by grinding a hole in Aluminum & fill it with solder, I think the mod is making the juntion thermoconductivity worse.

Click to expand...

I agree, if he wants to solder the LED directly to copper heatspreader, he's making life difficult by using a star.

Just solder the LED directly to a pedestal on the copper and solder wires to the LED pads.

There are quite a few examples on cpf.

Direct surface mount and soldering wires directly to leads on emitter sounds reasonable, although I've tried and failed miserably several times. Perhaps using different solders with different melting temps for the surface mounting and the wiring would help, because every time I try it I either melt the wires or the emitter loose. I do have another idea I'm working on to more closely simulate direct surface mounting without having to struggle with soldering wires to the emitter, or using a modified mcpcb. I'm going to hold off on describing it until I see if it is actually easier than surface mounting, and to make sure the emitter works after I'm done with it.