Heat sinking - minimal designs have enough??
I have been very impressed with the recent apperance of minimalist designs. I am in the process of building my first light and am concerned about the heat sinking capabilities of some of these minimal designs.
I know as the temp gets hotter, things tend to melt and efficency ultimately goes down. How hot are peoples projects running and when is it too hot?
I was finally able to run my design the other day. It's shown below, as posted on page 3 of Quazzle's thread: A nice alternative to P7 LED (1100 lumens guaranteed)
It gets warm pretty fast. I havn't been able to take temp measurements, but by touch it feels hot. I will be lengthening the housing and making deeper slots to add more surface area, but it will still be quite small. I'll be able to almost double the surface area with a length of 1.5" (as shown it's .9" long).
I do believe quite a few light designs on here are pushing the thermal bounderies closely
specially as the led rating has gone up some of mine included .
not too much of an issue in the UK but I think for you luckier folks with warm night riding may have the odd overtemp situation .
small is good but you may not be able to use the max power but then the rewards in battery runtime of running lower drive currents is a bonus
Totally agree. The lights I have built run flawlessly on colder nights but the triple xpg single xpe @ 1amp will get hot enough to trip the thermal management (60ºc) on a hot night ride if not going fast enough. Think I used the 1sq inch per watt rule. Anodizing black to improve emissivity would certainly help, or build a winter and summer housing.
Originally Posted by troutie-mtb
Mine is running very close to 18"sq surface area despite the small size. Standing still it gets hot, but when moving it never gets more than warm. Admittedly it is winter here at the moment, so I'm riding in sub 10 degree C temps right now. Even so, I think it would only be a problem if I was riding very slowly in warmer temperatures. But riding slowly wouldn't really require the full grunt. I wouldn't go any smaller than my current design.
Posting on the basis that ignorance shared is ignorance doubled.
which is exactly right, I have 3 power levels on all my lights, low for uphill, medium for cruising and high for screaming downhill. Only on high will the lights get hot and only if not moving
If its got thin flat bottomed fins, its OK
I like the 1 square inch per watt rule. It is simple and seems to be the absolute safe minimum to ensure appropriate thermal management.
The problem is how many watts does a particular emitter/LED use?
For example, I am planning on building my project around a L33.2MC from Quazzle. How many square inches of cooling would you guys recommend?
Update. Just went to the park to do some beam shots. Left the bike at home. Walking pace with lights on full at 5 degrees C the heads merely got warm. It would be a different story on a warm night I guess.
Posting on the basis that ignorance shared is ignorance doubled.
Ive been experimenting with Troute style handlebar sinked designs. They work, as long as you use an aluminum bar. Efficiency isn't so important with this design since the heat sink is so powerful. In theory it can use the bar, brake levers, stem and headset- maybe even the frame and fork's mass and surface area for cooling. I estimate you could cool at least 100 Watts of heat with no added heat sinks.
Ive built a prototype already and it cools a single XRE fine, of course. I just need to get some 20mm optics to focus it. I like using copper since its cheap, easy to work with and conducts heat better than anything. The next one im going to make will use a single or triple 20mm XPG setup. I'm planning on using the light with a dynamo too, so the driver is with the battery pack and the switch is a remote handlebar mounted one.
The only problem that I can think of is with the solder used to bond the bar mount to the flat piece the led sits on. Its made of lead, which cant conduct heat very well. The copper parts all touch though so it shouldn't affect heat transfer too badly.
Here are a few of my 3d sketches. The thermal path is only a few millimeters.
Last edited by Uzzi_PA; 08-10-2010 at 05:07 AM.
Using the handlebar and also included would be the stem is a great idea to increase the surface area needed to heatsink a light. The only issue I can see with the designs in the picture is the light would be handlebar size specific and the metal on metal may damage the finish the handlebar(if that is actually a big deal).
Another idea is the minimalist body screwed to a handlebar adapter plate. This way if could be changed out depending on the size of the handlebar.
You could make the design with two tabs on opposite sides such that a large rubber O-ring would hold the light onto the handlebar to further reduce the complexity. If it is light enough, this method should keep it nice and steady.
This has the potential to become the most minimalist light yet at the same time one of the best heat managed.
Now we have to see someone make one!
I'm no engineer, but I thought that each time you step from one surface to another, you lose thermal transfer efficiency regardless of the thermal grease/adhesive you use. In other words, going from die to MCPCB, MCPCB to housing, housing to air...each step dings thermal transfer. Now this is a bit off topic, but does a larger MCPCB mounting for a given LED improve thermal transfer to the housing? Any practical difference between the heat shedding properties of a 10mm and a 20mm or bigger star as far as the die is concerned?
Somewhere, several years ago, I heard 2 sq. in. per watt was a good rule of thumb to follow. I have used that or near that on my lights with good results. Yes, at room temp in still air they will reach 50-60C in about 5 minutes on max. A walking pace will keep them easily below 50C in warm night air of ~18-20C. ~8C air keeps them just lukewarm. You can get a pretty small housing to work if you can get a decent amount of fins on it. Look at this thread for a pretty small light that still manages heat well.
Micro helmet light
Agree with 1 sq. in. per watt rule of thumb. More if hotter/slower less if cooler/faster.
The unfinned Marwi bullet light with triple XPG's at 1 A is about 0.7 sq. in. per watt. The two lights have different copper spreader/firewalls, larger in the bar light. Designed with a 10 W, not 15 W thermal load in mind.
With some time now at 90-95 F (32-33 C) day temps and night temps of 80-85 F (29-30 C)and repeated checks by feel of units' temps, I found:
-In the day, with the helmet light on 1 A and flashing with stop lights, signs, slow traffic etc., when I have checked after riding slow or being stopped awhile, it it is warm but not hot, about hot tub temp.
-The bar light is run at 0.5 A constant mode day or night, and feels only slightly above ambient, a bit more in the day, but being black in full sun, it should be a bit above ambient if not gettting much air flow. That would be 1.4-1.5 sq. in. of radiating surface per watt.
The output from 1 A is overkill for most of my riding. The reflection back off signs can affect my night vision. So I get the runtime/temp benefit that Troutie mentioned. However running 1 A on warm nights just to see how the temps go, I found:
- If at about 15 mph they feel ambient or a bit warmer, faster and they are ambient or close, slower a bit warmer.
- If I stop and don't turn them down, I have had them hit the Bflexes' temp setting of 50 C and dropping to half output (Oh! I forgot to turn them down= dumb slap).
So the observations match the 1 sq. in. per watt rule of thumb for a light without the copper heat spreader as a design minimun and 1.4-1.5 as a reasonable maximum unless it is very hot or for a very slow bike.
As to having a separate winter light: winter temp tests showed them to run about 20 C above ambient (measured this time) when standing for over 5 minutes and stabilizing there, but they felt as cold as the bike frame at speed. That would put them at 0-20 C on most winter rides here so you get more light (CREE doesn't plot lower temps for these LEDs, but for some other LEDs they do, and the line is still straight at lower temps), assuming the oputput line continues, maybe as much as 10% more. A small weight and size penalty for a bit more light and simplicity of a one light system setup.
It could be argued a light housing (most likely doubles as a heat sink) should be designed for the hottest evening temperatures. If 1 sq inch per watt is the minimum, it could easily be argued 2 sq inches per watt is ideal.
I have ridden in times when the temp has pushed 80 degrees at night with minimal air movement on hills.
Back to my old question - anyone know how many watts the L33.2MC puts out???
3 XP-Gs at 1300mA (seems to me I remember Quazzle was running at 1300mA) will be 12.5W add 10% as a rough factor for the driver =13.8W.
Originally Posted by fujio001
Quaz is running 1.1amp
Mine stay around 50-65C on warm nights on high at nl riding speeds.
Last edited by chelboed; 08-10-2010 at 04:44 PM.
He has a new/different version so maybe both current set-ups are possible.
Originally Posted by quazzle
Yes that is my understanding too. Getting the heat out of the MCPCB into the light body is a separate issue from the radiative area to the air. One is part of the resistance in the heat path (like a pipe) from the LED to the interface with the air, the other is the rate of flow out to the air (like a spigot).
Originally Posted by buddhak
Yes, my understanding also is that a light body with an extra interface will have more resistance at that interface than had it been solid metal for the SAME AREA of contact. BUT if the area is large enough it won't pose a significant thermal barrier as there is a diminishing return with larger thermal paths in solid metal. A large mediocre interface is soon as good as solid metal. Sort of like a pail with a lot of holes in the bottom moving as much water or more than an inverted water bottle both with the same height (pressure) of water in them. If that's all you need to move to keep ahead of the inflow, they are equal. One is many parallel streams like an interface the other a solid stream like solid metal.
One of the reasons for MCPCBs is to spread the thermal load to a larger area because the interface is not as effective thermal medium as solid metal. That said, a larger MCPCB may or may not allow an LED to run cooler. If a smaller one and the heat sink are polished to get more than 4 times the metal-metal contact, it may provide a better thermal path. So high power LEDs on small MCPCB's will respond well to best mounting practices or the converse, be hurt more by a bit too much thermal paste or rough surfaces than larger ones.
Since the thermal resistances are additive, you would think that they would be more important in a small light where the air-light interface may run hotter. They are BUT the resistance at the LED junction to pad interface is at least an order of magnitude higher so it dominates total thermal resistance from the junction to the air interface. The air intercface resistance is in the same order of magnitude. That means as long as the resistance at the interface is small compared to the LEDs internal resistance or resistance at the air interface, a working for a bit smaller resistance helps only a little bit.
Since we don't have infintite mass unless we use an alumium handlebar and the rest of the bike, it is the dissipation of heat to air that is essential (AND the topic of this thread. See? On Topic!). In other words, if you have a light with inadequate air interface, or if you stop moving a light that has enough surface with air movement, or if the air temp soars, then the flow of heat is reduced, and the light gets hot as does the LED. The small internal resistances provided they ARE small, play little role. The air interface is the issue. In the case of the Magicshine MJ808, the thermal path is so tortuous that the thermal resistance of the heatsink interface to the light body is large enough to play a role in its LED overheating, so some care is required to keep the resistances minor with decent design and good interfaces. (The MJ-808 is also pushing the optimal area per watt pretty hard.)
Testing a light with triple XP-Gs @ 1A and 27 C ambient temp, and protected from drafts, the temperature differential from the MCPCB at the LEDs (as close as I could get), and the body was no more than 2 degrees C with the light body at 20 degrees C above ambient. All interfaces were polished and thermal paste was used and the MCPCB and heat sink were lapped. That was overkill. The distance of the thermal path through the metal explains most of the 2 degree temperature difference. It is hard to know when it is good enough, though if you aren't building and testing many of the same design. Nothing excells like excess!
This thread is bringing up some very good points in buiding my light. In addition to having enough surface area for cooling the interface between metal pieces can make a difference. I remember reading about this when I was building a computer.
So my L33.2MC requires about 14 sq in to cool at a safe minimum. The hammond box I am using is barely adequate.
Now the point of metal to metal interface. Since the front back plates bolt on and I have to make it waterproof, it is effectively not part of the heatsink.
In addition, if possible I need to optimize the interface between the LED and aluminum angle piece. The challenge is going to be how you maximize thermal transfer from the angle aluminum and the box itself. Both surfaces are not exactly smooth and I can imagine thermal transfer will be hurting.
Lots of things to consider when building a light. There is more than just slapping it together!
Now I understand the superiority of maching a housing from one piece of aluminum with lots of fins acting as heatsinks. Only one heat transfer you need to worry about - LED to aluminum housing.
Ah, MTBR. For every troll there are 2-3 teachers. Thanks BrianMc.
They will radiate what heat does get to them through the fasteners. More screws? You could add a seating area, or brackets inside of the area of the seals to connect them better thermally to the case. That will be easier for the back than the front.
Originally Posted by fujio001
One square inch per watt is minimal only if you run on high at slow speeds or stopped a lot. I am getting away with 0.7. Part of that may be a short and good thermal path to the body. So I think the 1" has some fudge in it for thermal paths that are less than ideal.
[QUOTE=fujio001] In addition, if possible I need to optimize the interface between the LED and aluminum angle piece. The challenge is going to be how you maximize thermal transfer from the angle aluminum and the box itself. Both surfaces are not exactly smooth and I can imagine thermal transfer will be hurting.[QUOTE=fujio001]
Some ideas for what they are worth:
Both contact surfaces of the angle are outside surfaces easily polished and lapped. Google computer heatsink and lapping for more details. I used plate glass (very flat) and worked my way up in abrasive number. The backs of 10 mm MCPCB's are hard to do without rounding the edges. Larger ones are a bit easier. It is the inside bottom of the case that is hard. You could use a wood block that you first planed flat and wrap the sand paper around it You can also use an angle (if it fits) that is longer on the side fastening to the case to increase the case-angle contact area making more microscopic metal-metal contacts. In fact, it could cover the rest of the bottom of the case and the area would be so large that with a few screws and thermal compund, you'd be fine, forget sanding. You can use mechanical fasteners to clamp the angle hard to the bottom which mashes the high points making them bigger, pushes 'almost contact points' together, and squeezes out excess thermal paste. Tighten from the center out to squeeze out the comound best. I used very small hobby screws to fasten the 10 mm MCPCBs in my lights to do the same thing. If you do a decent job, the distance through metal from the LED to the air interface becomes a bigger factor in the thermal path resistance than the resistance at the interfaces of the parts. If the case is flat bottomed, and you have clearance either side of the handlebar mount, you could use the same fasteners to clamp aluminum channel or IC heat sinks to add surface area right under the angle aluminum and improve two issues, thermal path and air contact area. Since those are out of sight on the bottom, the lights appearance isn't hurt much either, should they prove a little ugly looking when done
Plumbing and electrical wiring are also simple, but if they are critical like on the Space Shuttle, it's more than just slapping them together. There is a simplicity and elegance of a purpose made machined housing, which is a kind of superiority, but I don't know as I would call them definitely superior in function. The proof is in the functioning and how easy it is for you to build it, and if it works as well. With diligence, you can match performance and close counts here, because enough thermal path is enough thermal path, anything more is wasted. A fire hose is excessive if you only need a garden hose. Part of the elegance of multi-piece designs is their being possible with benchtop tools and a somewhat competent tool user. The "I can do that!" factor is it's own kind of elegance.
Originally Posted by fujio001
Love to see your finished light.
Very well said. i will take your advice seriously in the construction of my light. I have almost all the components I will need.
The housing will be a Hammond MFG waterproof case - powder coated black (from the factory), Judco switch, Hammond waterproof cable gland, Mawri mount (Thanks Hoffman Amps!) and a Magicshine battery with waterproof casing. I will attempt to use on the extension cords for the magicehine as all the wiring I will need. Also have the artic alumina adheisve, thermal compound and some thermal tape.
I also am waiting for a variety of other adhesive to include superglue gel and silicone adhesive.
The final component which should be here in a week or two is the L33.2MC. I will only use one on my first build. I think 1100 lumen is a good place to start. We will see.
Even before finishing this project I am already dreaming of a enclosure which will house 2 of the L33.2MC. 2200 lumens!
I should have time to get some of holes drilled into the case tommorow.
Reawakening this thread .
When you say 1 sq inc surface area per watt of power how are you calculating how many watts the light housing has to disipate .
is it worked out from the Volt and amps going in to the light from the battery .
like 15 volts @ 1.5 amps going in = 22.5 watts and we know a very small % of that is converted to light so that light should be 22.5 square inche of surface area .
Yes Troutie, but the light out is a reasonable portion, I'd call it 25% (after the driver).
Originally Posted by troutie-mtb
Is it warm?
Nah, not really.
She'll be right