Imalent        
Page 1 of 2 12 LastLast
Results 1 to 30 of 50

Thread: Heat Management Issues

  1. #1

    Default Heat Management Issues

    I've lurked around here for the last couple years. Recently I have had a bit more time that I can spend to move beyond bolt-on mods. There's alot to learn on the diode and electronics side and I've begun to ask some questions. I thought I might give a little back in the exchange.

    In my "day job" I deal pretty regularly with heat management. I've picked up a bit of experience over the years that might be helpful to some board members. Before writing this, I ran some searches here and found that while the topic comes up occasionally, there are alot if points missing in the dialog.

    As I read through threads and examine the various builds, it's very obvious that the heat generated by these various components when pushed to their limits becomes substantial- and often times a barrier. In more than a couple builds, I see things being done that while logic might lead one to believe might help in the management of heat, they are actually detrimental. I hope some of this information is helpful.

    Heat-
    Understanding the barriers to good thermal management takes a basic understanding of the forms of heat and how it's moving. In the case of a flashlight, you have two things going on. Conductance and emittance. In simple terms, when you heat an object, you excite its molecules. As they crash into each other it creates heat. That heat moves through the mass of our piece of metal as conductance. Conductance is the energy of one molecule transferring it's energy to the one next to it. This is best exhibited by heating the end of a metal rod. When you start, and if you are applying enough heat, the end you're heating get's hot quickly. If you bring it to a temperature and hold it there, the other end of the rod will begin to get hot. I'm guessing everyone here has experienced this with a bench grinder or sweating a copper pipe. This is called equilibrium. Eventually every molecule in that rod will be at the same temperature. This is only going to happen in the real world if you do this experiment in a vacuum.

    That brings us to our next facet which is emittance. This is primarily what you're feeling when you hold your hand close to the end of that heated metal rod. This is no longer the transfer of energy at the molecular level. This is the conversion of that energy into electromagnetic radiation. The hotter you get the metal, the higher the frequency of radiation. That's why at lower temperatures that piece of steel is cherry red and at some point it becomes white. Again, in our real world, we are experiencing both. If you hold your hand close enough, part of what you'll feel is the conductance of heat through the air. But air isn't a very good conductor so you don't have to pull back far to reduce that conductance dramatically. This is all going on inside of your flashlight every time you turn it on. And there are simple things that will allow it to operate at better efficiency.

    Materials-
    The preponderance of these custom builds and mods are builds done in aluminum. Some of the higher end stuff integrates copper. A few are done in titanium. Each of these metals have obvious benefits and drawbacks- and a few maybe not so obvious.

    Aluminum is a wonderful material in that it has a strong strength to weight ratio, it's relatively easy to come by inexpensively. But one of the best benefits to aluminum is it's corrosion resistance. It's corrosion resistance comes from aluminum's propensity of rapid corrosion. Aluminum exposed to oxygen begins to oxidize very quickly. This reaction creates aluminum oxide- a crystalline ceramic. After a very thin film forms (a couple nanometers) the oxide seals the surface, preventing further oxidization. It takes more than the relatively low reactivity of the air in our atmosphere to penetrate that coating. Acids and bases (salt) certainly cause continued oxidization. Apply a little electricity and a few of these other elements and you have anodizing. Here is where our "problem" surfaces. Aluminum oxide is a very poor thermal conductor. The average layer of oxidization in the anodization on a flashlight body can create a barrier in the magnitude of 10 to 30 times that of bare aluminum. Anodizing is a very effective thermal conductance inhibitor. That's bad. But not always. Anodizing is one of the most thermally emissive materials. It's about as close to black-body as you can get. And that's good. That's why alot of heat sinks are anodized (color doesn't matter- black-body is just the thermodynamic term). So inside your flashlight, if that new heat sink you just machined is fitted into an anodized body, it has the equivalent of a thermal blanket in between. Even a piece of bare aluminum isn't much better unless it's had the layer of oxidization removed before fitting the parts. And even then there are steps necessary to prevent future oxidization. The outside being anodized- good thing.

    Copper is another material I have seen used. Let's look at copper as if it were a highway. Our copper highway is four lanes wide. A little traffic moves absolutely unimpeded. So does moderate traffic. Heavy traffic moves right along pretty well. But if there's only one toll booth open at the end, it could be 30 lanes wide and eventually it will fill up. Our one-lane aluminum highway with one toll booth open is letting just as many cars out as 4 lanes of copper. Copper will transfer heat rapidly but if it isn't able to disperse that heat, once the mass is saturated (equilibrium) it has no benefits beyond aluminum. In a flashlight design, if you haven't figured out how to allow the heat to conduct away from the mass, it really doesn't matter how fast you are initially moving heat as it's going to slow down pretty soon. If you want to run your light for a few minutes, you'll never see a problem. If you want your light to weigh 23 pounds, you'll never see a problem. But if you want a small, light flashlight that will run for extended periods, you have to move heat. Copper and aluminum react with each other and each material oxidizes so unless addressed you have the same issues as your aluminum-aluminum interface.

    Titanium is about as cool as it gets in obtainable metals. I have a Ti paperweight. There's a reason Lockheed built the mach 3-plus SR-71 out of titanium. It barely conducts heat. Ti is great for hypersonic aircraft and hip implants but not so good for high-output flashlights.

    Other metals- If copper is so good, brass must be as well, right? I mean, it's mostly copper. Brass is three times less thermally conductive than copper and about two times less than aluminum. Silver is very good- about a few percent better than copper. Both oxidize about the same. The best conductor? Diamond. At the extreme edge of thermodynamic management we're starting to see diamond heat sinks. Back to metals, you have to be careful in your selection. Anything alloyed with copper significantly impeded its thermal abilities. 6061 aluminum is better than 2024. 0 temper 6061 is a bit better that T6. Not enough to worry about. But all things being equal, you might as well get what you can on your side.

    Thermal management strategies-
    In talking about the thermal conductivity of metals, I missed one. Lead. Lead is a terrible thermal conductor About 1/1oth of copper ans 1/6th of aluminum. If you used common solder paste to affix your diode to that heat sink, you placed a huge barrier in the way of it's heat conductance. Semiconductor manufacturers are now turning to exotic metals to aid in thermal management. One of the most attractive is indium. Done in a similar fashion to furnace brazing, a small piece of indium sheet is placed between the component and the board and the piece is heated. The best part is it melts at about 100c. You could do that with a good hair dryer. Indium is 3 or 4 times more conductive than lead. Here's one that really shocks people- thermal paste is almost worthless as a thermal conductor. Someone showed me once a test of most of the commercially available pastes and ordinary zinc oxide and common grease worked just about as well. One compound works astonishingly well and that's gallium. Gallium is a metal that, like mercury, exists in liquid form at room temperatures. Downside- it's highly reactive with aluminum.. Indium and gallium are so much better because they are molecularly much more similar to the metals used in these components than grease is. We're now looking at this from a mechanical perspective at the micron level. The reason that block of aluminum in your hand doesn't separate into its gazillion separate atoms and disperse into the wind is molecular attraction. If you were to take two blocks of that aluminum and perfectly microfinish the surfaces absolutely flat, touching the two faces together would make one block. They would be inseparable. This isn't possible with our present level of technology. But I've seen something close. It was a laboratory grade surface plate with gage blocks sitting on it. They were so flat and so polished that couldn't be lifted off. They had to be slid to the edge to remove them. This is why these metals work so well. They're bonding at the molecular level. I went through all of this to illustrate a point. If you machine a pill and slide it into an anodized aluminum tube, you're leaving alot on the table. Removing the anodized layer and using thermal compound isn't much better. This is why all the instructions you see tell you to use the absolute thinnest amounts of thermal compounds and adhesives. You really want as much of the metal you can to be in contact. The thermal compounds just fill the voids. So take that heat sink and polish it to the finest possible point that you can. The tube it's going into as well. Then join them in an interference fit- press them together. Your thermal transfer is now many, many times better than a 0.001" gap filled with Arctic Silver. This goes back to our differences between conductance and emittance. A rough finish viewed under a microscope would show that the heat sink is only making physical contact in not alot of places. These are the only points that conductance is occurring. In all the other spots, emittance is at work, and that's way more inefficient. Air is a lousy thermal conductor. I can't make any recommendation on this yet but I've begun to look into aluminum soldering. Ideally our assembly is made of one piece of metal. The diode is on the star with something like indium and the star is bonded to the body in the same fashion. This is out of reach for all but the most advanced fabricators. Aluminum soldering seems to me to be the ideal fashion of bonding a sink into a body. In practice, the heat may be too high. You could certainly do it with a thin sheet of indium. For a few hundred dollars.

    To summarize-

    -Oxides and surface finishes conduct a fraction of what bare metals do
    -Alloys typically conduct much less than pure materials, some significantly more
    -Metal-to-metal contact is always the best thermal pat
    -Avoid thermal compounds, apply very thinly when used

    Alot of this is unimportant in the majority of flashlights. But in the short time I've been around the build and mod side, the more signs I am seeing that these issues will become more prevalent. Batteries are getting smaller while at the same time they are seeing higher current and voltage outputs. Led size is shrinking while at the same time output is rising. Inevitably, technology on the manufacturer side will bring more efficient diodes that generate more light and less heat. In the interim, some of this may help you.
    Last edited by CTS; 11-29-2012 at 02:28 PM. Reason: Wasn't quite long enough

  2. #2
    *Flashaholic* Gunner12's Avatar
    Join Date
    Dec 2006
    Location
    Bay Area, CA
    Posts
    10,064

    Default Re: Heat Management Issues

    Great post about an oft overlooked subject!

    LED die size seems to have increased, the Luxeon III and XR-Es had 1mm x 1mm dies, but the more common XP-G and XM-L have 1.4mm x 1.4mm dies and 2mm x 2mm dies respectively. The package size has decreased, but the die size seems to have increased.

    How would a thin layer of solder do compared to indium?
    How much worse is a screwed in pill compared with a good interference fit? I know it'll depend on how tight the tolerances are and how smooth the threads are, but overall, do you have a ballpark figure of the difference in thermal conductivity?

    Thanks for the informative post!

    And if I haven't said it already:


  3. #3

    Default Re: Heat Management Issues

    Quote Originally Posted by Gunner12 View Post
    Great post about an oft overlooked subject!

    LED die size seems to have increased, the Luxeon III and XR-Es had 1mm x 1mm dies, but the more common XP-G and XM-L have 1.4mm x 1.4mm dies and 2mm x 2mm dies respectively. The package size has decreased, but the die size seems to have increased.

    How would a thin layer of solder do compared to indium?
    How much worse is a screwed in pill compared with a good interference fit? I know it'll depend on how tight the tolerances are and how smooth the threads are, but overall, do you have a ballpark figure of the difference in thermal conductivity?

    Thanks for the informative post!

    And if I haven't said it already:

    Depends on the solder. Somewhere around here I have a comparison of various commercially available solders. The best one thermally is about 97% tin and 3% silver. Some types are as much as 20% conductive thermally. Indium is exceptionally conductive, beyond all all the other materials. But it's disproportionally priced for its return. And certainly nothing a casual hobbyist could easily work with.

    I'm not a metallurgist so I can't explain why, but different blends of metals have often times some very dramatic thermal characteristics. A few percentage points can completely alter the way the metal conducts heat.

    The most precision-cut thread isn't going to approach an interference fit- provided that you get the surfaces of that interference fit are as smooth as you can possibly get. Imagine you're a piece of dust and you're standing on a sheet of 120 grit sandpaper. You're looking up at some very tall and very pointy mountains. Now turn that piece of sandpaper upside down and lay it on a sheet of glass- like a mirror. Imagine what you'd see if you were that piece of dust. Only the very tips of each of those mountains would be touching the glass. Everything around them is air- probably 95% or so of the surface area. And air is a lousy thermal conductor. Now start applying pressure to the back of the sandpaper. At some point, the mountains are going to be crushed down and parts of them will be pressed down into the surface of the glass. Not alot of air in there. Think now of two mirrors laying on top of each other. Again, very little air in there. Keep in mind that our sandpaper is a sheet of Mt Everest's compared to our present ability to surface finish in the micron range. And that finish is a whole sheet of Mt Everest's compared to what you would see if you could see the individual molecules. For heat to efficiently transfer between molecules, they have to be touching. If there's not, there's something in between them. And if not that, it's a vacuum, which transfers no conductive heat.

    Back to threads- one of the first obstacles is that the act of threading itself pulls the surface contact to one side of the thread surface interface and away from the other. Now you have surface finish. The best possible threading process is going to leave an incredible amount of surface imperfections. And other than electrochemical polishing there's no real way to fix this. ECP would still not address the "fit" of one surface to the other. A smooth surface gets you part way there. An interference fit crushes some of those remaining "mountains". Relatively speaking, solder is better. I'd be curious to experiment with a very closely matched taper fit, bonded with a high-metals bonding agent. They could be pressed very firmly together during the cure. Eventually, someone will come up with an indium filled material. Possibly that's been experimented with already. If you want to skip all this, find a good local welder and have them weld the interface. Probably impractical- you'd at a minimum destroy whatever finish it had on it.

    To put a number between the best thread a hobbyist machinist could likely do compared to an interference fit? 40-50%.

    I feel the need to emphasize this again- we're talking about things that are only going to matter at the very most extreme point of performance.
    Last edited by CTS; 11-29-2012 at 06:26 PM.

  4. #4
    *Flashaholic* Gunner12's Avatar
    Join Date
    Dec 2006
    Location
    Bay Area, CA
    Posts
    10,064

    Default Re: Heat Management Issues

    I've been planning on making some lights for a while now, but I'd also like to keep it easily upgrade-able, so a threaded pill would be much easier to remove then a interference fit. Thanks for the ballpark figure. So 10-20% might be a more realistic figure of thread vs interference fit? Also, how much would some thermal paste in the threads help the thermal transfer?

    I've seen some fine polishing wheels before, 0.1 micron in a lab. I think they were using it to polish and inspect pieces of metal. The professor brought in a ceramic that with a little bit of water (and a smooth finish), will bond to itself very well. The two pieces he had had to be slid apart. There might have been some hydrogen bonding happening too, but it was pretty cool to see.

    I was looking up thermally conductive solder too a few months back, here's the first site that popped up on google: http://www.electronics-cooling.com/2...ty-of-solders/
    The next link is also pretty nice, and also lists the thermal conductivity of indium as well: http://alasir.com/reference/solder_alloys/

    Silver solder is the best thermally, 91/9 Tin + Zinc solder also does pretty well.

    If threads were matched perfectly, the metals would bind, making it near impossible to remove anyways, thus defeating my reason for threading in the first place :P

    Thanks for the detailed reply CTS

  5. #5

    Default Re: Heat Management Issues

    This topic jogged my memory on something I had seen a few weeks ago. I'm building my first light using the XP-G. I downloaded some docs from Cree. The attached PDF is one on thermal management. I suppose I didn't really consider just how much a barrier the PCB presented. Cree's stated specs are most likely based upon their own modeling of thermal performance as viewed as a total package. In essence, when they're rating a diode, they're looking at its ability to perform to spec as placed in a system and not as if the diode was being examined individually. What I'm getting at is this- looking at the PDF, it shows the thermal limitation of a glass PCB vs a MCPCB. It then goes on to discuss the additions of what they call "via's" which are channels oriented vertically through the MCPCB. They're using these in the suggested implementation being filled with solder for the purposes of channeling heat. Looking at this instantly made me realize what a thermal bottleneck exists at the MCPCB. That tells me that most likely the diode could be driven significantly beyond rated spec if this obstacle could be overcome. It wouldn't be all that terribly hard to accomplish.

    http://www.cree.com/~/media/Files/Cr...CB_Thermal.pdf

  6. #6

    Default Re: Heat Management Issues

    Quote Originally Posted by Gunner12 View Post
    I've been planning on making some lights for a while now, but I'd also like to keep it easily upgrade-able, so a threaded pill would be much easier to remove then a interference fit. Thanks for the ballpark figure. So 10-20% might be a more realistic figure of thread vs interference fit? Also, how much would some thermal paste in the threads help the thermal transfer?
    I believe the difference would be higher.

    I posted on another thread started just about the same time I was writing this one. It was concerning thermal pastes and adhesives. One manufacturer claimed a very high conductivity number. It reminded me of a study done a few years ago of thermal adhesives. They ranged from el-cheapo impregnated RTV silicone to high silver content polymer. At one point, the silver blew the RTV away. But that was at about half a millimeter thick applications. At about 10 thousandths, the materials performed about the same. In this test, their interface materials were microfifnshed aluminum blocks.

    If I were you, I would skip the threads and focus on surface finish. If you left a small space in between that you could use a good thermal paste to fill the void, I believe that would be more beneficial than trying to close the tolerances of threads. I would also look at physically retaining the heat sink by using the bezel/head to clamp it down into the body tube, if possible. Maybe use some very fine lapping compound to mate the faces and seat them together as closely as possible.

  7. #7
    *Flashaholic* Gunner12's Avatar
    Join Date
    Dec 2006
    Location
    Bay Area, CA
    Posts
    10,064

    Default Re: Heat Management Issues

    Or maybe have two screws secure two well polished surfaces, kinda like how the DEFT was built (sabbluster's light). One would be the "pill", the other will be the interface to the light. Those surfaces will mate, creating a good thermal path.

    I also missed that thermal PDF on Cree's site, reading over it right now.

    Some of the MCPCBs are pretty crappy at transferring heat. There are ones where the metal core is copper, and the thermal pad is a hold directly to the metal core, thus decreasing the thermal resistance by removing the dielectric.

  8. #8

    Default Re: Heat Management Issues

    Obviously what we're talking about here is extremely small interfaces. If your threaded area is the size of the ID of a 20mm tube and 10 mm long, that area is going to conduct plenty of heat. If your thread is 3mm, that's another story.

    If you visualize an MCPCB, right below the top copper face is a thin film of dielectric. The solder joint is conducting at 60, the copper is conducting at 400 and the dielectric is conducting at 1.5. While the 1.5 looks like the barrier, it's conducting 1.5 through 310 sq/mm (20mm board) 1.5 x 3.1 is 4.65. The diode is connected with .04 sq/cm of solder at 60. 60 x .04 is 2.4. Cree is rating the diode on a glass board with 50% the conductivity, which would match right up with the thermal conduction ability of the solder joint. Clearly the Achilles's heel in this scenario is the solder joint. If you were to use 50/50 solder at this joint you would lose 2/3rds of your conductivity. That's bad. On the other side of the MCPCB, whatever you use to bond it doesn't need to have a thermal conductivity beyond 1.5 since that's the max the board is going to be able to transfer into it.

    I have no idea what the heat output curves look like at various drive levels. I'm assuming since the efficiency curves I've seen all represent reduced lm/w output as levels increase, that wasted wattage is leaving in the form of heat. One might be able to extrapolate from that just how far a diode can be pushed until the thermal conduction of the solder joint is saturated.
    Last edited by CTS; 11-30-2012 at 06:49 AM.

  9. #9
    Flashaholic* The_Driver's Avatar
    Join Date
    Jan 2010
    Location
    Germany
    Posts
    1,165

    Default Re: Heat Management Issues

    @CTS: please take a look a the measurements in this post.
    The Varapower 2000 is a heavily modded maglite that uses an overdriven Luminus SST-90 led (up to 15A!!!). It has a massive aluminium/copper heatsink which is intereference-fitted into the anodized aluminium maglite body. In this test the 2 lights were identical except for the heatsink material. In both cases the LEDs were mounted on their normal star PCBs and then boltet onto the heatsinks with thermal paste in between (see picture here).
    The measured numbers are clear and cannot be explained by the usual led variances.
    Also note that the lights are not really regulated, the current falls at the beginning due to the battery voltage sagging under the extreme load.

    These days the copper heatsinks in the current Varapower Turbo lights (for pictures see my translated review here) are 20% larger and the emitters are directly soldered to the heatsinks. No otf measurements have ever been done, but the lux readings suggest that the lumens are easily over 2000 otf (out the front).
    Last edited by The_Driver; 08-26-2018 at 03:42 PM.

  10. #10

    Default Re: Heat Management Issues

    Another great link that's appreciated.


    What we're talking about here is the numbers side- the engineering assumptions based on known quantities. The kind of stuff you'd get from a new BSME grad. Ask him 15 or 20 years later and he'll tell you that all of that is a really good guideline for where to begin a design. I also didn't really get into anything but basic stuff. Things like relatives of efficiency aren't even in the mix. So I would definitely expect the physical results to vary- even between identical units.

    It's going to take some pondering to come up with an explanation on the lambda. My initial thinking is that even though he is using a copper board, it's pretty small in diameter. His design might have a thermal architecture that puts its conductivity just slightly beyond what that small area of interface could accommodate. If I were to take a wild guess, I'd say a larger PCB would do the same thing on the aluminum sink. If you were to bond that diode directly to an aluminum post the exact same area, 9sq/mm, the aluminum sink would work significantly better. It doesn't matter if it's aluminum or copper, the interface is only capable of carrying 60w/m K. Aluminum conducting at close to 300 or copper conducting around 400 exceeds the bond joint by a factor of 5+. If I could pull that assembly apart, I think I would find a relatively small conductive path from diode's die back through the board and to the sink. Just enough for it to work on copper and just not enough for it to work on aluminum. In this case the copper sink allows the designer to get the diode out and into the reflector. In this case, copper does present an advantage to the overall design.

    I would really love to be able to look at some empirical data from actual side-by-side testing that took into account all the variables. You would have to have thermocouple before and after each interface. It would certainly show where the bottlenecks are.

  11. #11
    Flashaholic* The_Driver's Avatar
    Join Date
    Jan 2010
    Location
    Germany
    Posts
    1,165

    Default Re: Heat Management Issues

    Quote Originally Posted by CTS View Post
    Another great link that's appreciated.


    What we're talking about here is the numbers side- the engineering assumptions based on known quantities. The kind of stuff you'd get from a new BSME grad. Ask him 15 or 20 years later and he'll tell you that all of that is a really good guideline for where to begin a design. I also didn't really get into anything but basic stuff. Things like relatives of efficiency aren't even in the mix. So I would definitely expect the physical results to vary- even between identical units.

    Ok, I guess most of us didn't study this stuff in university

    It's going to take some pondering to come up with an explanation on the lambda. My initial thinking is that even though he is using a copper board, it's pretty small in diameter. How do you know that he is using a copper "board" (=PCB?)? His design might have a thermal architecture that puts its conductivity just slightly beyond what that small area of interface could accommodate. If I were to take a wild guess, I'd say a larger PCB would do the same thing on the aluminum sink. If you were to bond that diode directly to an aluminum post the exact same area, 9sq/mm, the aluminum sink would work significantly better. It doesn't matter if it's aluminum or copper, the interface is only capable of carrying 60w/m K. Where did you get this number? Aluminum conducting at close to 300 or copper conducting around 400 exceeds the bond joint by a factor of 5+. If I could pull that assembly apart, I think I would find a relatively small conductive path from diode's die back through the board and to the sink. Just enough for it to work on copper and just not enough for it to work on aluminum. In this case the copper sink allows the designer to get the diode out and into the reflector. What do you mean with this sentence? The heatsinks are both identical in size and are manufactured specifically for these lights. In this case, copper does present an advantage to the overall design.

    I would really love to be able to look at some empirical data from actual side-by-side testing that took into account all the variables. You would have to have thermocouple before and after each interface. It would certainly show where the bottlenecks are.
    If you have all the required equipment available why don't you do a test like this? The entire community would benefit from your findings. You could take 2 common leds (like the Cree XM-L and the Luminus SST-90 for example) and measure everything you want.

    Also regarding the Varapower 2000/Turbo: how much difference do you think it makes when the LEDs are soldered directly to the copper heatsink?

  12. #12

    Default Re: Heat Management Issues

    No college for me. But 35 years of "doing" and a pathological desire to know everything I can about the things that interest me. I don't do hobbies. I have obsessions.

    To your questions-

    I'm assuming it's a copper board. It looks like one and in this application I couldn't imagine an accomplished builder would use anything but the most advantageous materials. Surely not for the difference of a couple dollars.

    60 w/m K what the best commonly available adhesion materials are capable of in real-world applications. Pretty much the best solder you can use is 97% tin and 3% silver. It's capable of conductance @ 60. There are some thermal adhesives out there rated at that level. One was referenced by the OP in the other thread. Conductivity above this is tough and expensive to attain. Conductivity at this level is as simple as using the correct material. 60 is about where the cost curve goes almost vertical, the effort curve about 60 degrees and the benefit curve about 5-10 degrees. But if your design needs just that one little pinch more, that effort and expense is probably what it's going to take. It's points like these where the designer starts asking "how bright do I want to spend?" In our example, a little bit more conductivity at the LED joint may have negated the need for the brass heat sink. If a filled polymer with elevated conductance were available, the designer could have likely stayed with aluminum. At $30, for a slug that size he's just about offering the upgrade at cost. There is a real-world takeaway from this discussion of esoterics- As I mentioned before, throwing away two-thirds of your thermal conductivity capacity is as simple as picking up the wrong spool. If you bought a bare XM-L and a blank board, then used 50/50 to solder the diode, I'm not sure it would run at spec sheet levels, let alone in elevated current states.

    On the heat sink question- I have read about the diode's need to be carefully positioned on its X axis. The position of this particular diode seems to require proper distancing to achieve the beam pattern desired. In order to accomplish this the mounting must be small in diameter as to keep the reflector opening as small as possible. If he would have put it on a 30mm board, he would have had to cut a 30mm+ hole in the reflector. Again, I'm a neophyte in these areas but my assumption is that would have negatively impacted the designer's goals. What I was trying to point out is somewhat of a contradiction. The design didn't benefit from that massive piece of copper at any point other than a very small area where it bonded to the MCPCB. I would bet that a copper disk no bigger than a nickle pressed into the aluminum sink would have accomplished the same thing.

    I don't have the equipment to do these tests. But at my current rate of increasing interest, I probably will in the near future. I'm sure some of these experiments would be quite fun.

    As far as the direct-connect to the sink- let's go with what we know about the test results you posted and use some numbers and assumptions that I think are definitely defensible...

    The build consists (we believe) of the diode mounted to a copper MCPCB @ approx 20mm and then affixed to the heat sink with thermal adhesive. Since we don't know what adhesive he used, it makes this tough. But if he used the Arctic Silver material, he's got 1.5W/m K of conductance over the surface area of 310 sq/mm. If you convert the 1.5 from meters to millimeters and multiply it by the surface area of the board, you get 0.00465 conductance. I believe most boards are capable of 2+ conductance. A 20mm board would have a conductance of 0.0062 so barring inefficiencies, our primary thermal obstacle is the bond between the board and the sink. Increasing board performance wouldn't do anything unless we can improve the bond at this joint. If you furnace soldered the board to the copper sink, you'd have something. But I don't know that the dielectric could handle that heat. You'd have to reflow the diode on as soon as you got the melt temp in the board/sink joint. Would it hurt the diode to be that hot for that long? The 9mm die of the diode will give us 0.0054 conductivity. Looking at these numbers, the board is able to conduct more heat out than the diode is capable of conducting into the board.

    For comparison, at 9 sq/mm you have a conductivity through copper at 0.036, aluminum at 0.027

    If you directly connected the diode to a sink with a projection of the same area as the diode's die @9 sq/mm with material capable of conducting at 60 W/m K, your results would be, using the same formula as above, 0.00540 conductance. That's a net gain of about 15%. This assumes you can bond the whole surface area with a non-conductive coating and connect current in another fashion. Maybe impractical/impossible? So we have to go to the part of the base that we know we can bond to- the center pad. Again, into an area where I have little knowledge, but if I understand correctly a Copper board has an area beneath the diode that is solid copper through the board. I don't have the numbers, but from looking, my estimation is that the base pad in the center of the die is about 60% of the total surface area. That's 5.4 sq/mm which gives you a conductivity of 0.00324.

    Clearly the use of a board is beneficial over a direct bond to the sink, simply because a direct sink bond allows heat sinking of the electrical contacts. And if my numbers are correct, the board is more conductive than the best bond achievable for soldering the die to the board. Reaching beyond the 60 W/m K level of typical solders is worth investigating.

    Let's toss into the mix a change in bond between the sink and the board. If this 60 conductivity truly performs at that level, or even at 10% of that level, the interface issue is now eliminated. The back side of the board now possesses 35 times the conductivity as the diode/board junction. Certainly easier than soldering the board to the sink.

    After looking at these numbers, I am thinking that all the copper is doing is delaying the inevitable onset of thermal performance impedance. The copper heat sink is simply more capable of absorbing and storing heat energy than aluminum. Once the material achieves equilibrium, heat stops moving. None of the components in the heat path conduct less than the thermal output of the diode except the bond between the board and the sink. If we assume they were bonded exactly the same, then I believe the thermal bottleneck exists at the junction between the heat sink and the flashlight's body. If he pressed the copper sink into an anodized tube, he's effectively wrapped a blanket around his heat dissipation efforts.

    Just a thought.

  13. #13
    Flashaholic* The_Driver's Avatar
    Join Date
    Jan 2010
    Location
    Germany
    Posts
    1,165

    Default Re: Heat Management Issues

    You write a loooooot (which is good though )

    Here is Kevin's (the maker of the Varapower lights) Website
    More importantly here is his sub-forum in a different flashlight forum where he talks about all his lights (I have also been active there).
    The prices of his lights are extremely reasonable for all the work that goes into them. Here you can see all the build steps required to make the older Varapower 2000 lights (these had stock maglite heads). The current Varapower Turbo lights have a much bigger head that requires a lot of extra work (it's made of a maglite head and a head from a much bigger light). He even makes the PWM-driver himself.

    Concerning the older lights form him: as can be seen in the picture I linked to the star-pcb is screwed onto the heatsink with a thin layer of thermal paste in between. Screwing it down with a lot of pressure is very important.

    Clearly the use of a board is beneficial over a direct bond to the sink, simply because a direct sink bond allows heat sinking of the electrical contacts.
    What do you mean with heatsinking of the electrical contacts?
    It is common belief in this forum that soldering an emitter directly to the heatisnk is better than any commonly used pcb-mounting-method. You only have one thermal barrier when it is directly soldered to the heatsink. A pcb adds at last one more to this.

    The copper heat sink is simply more capable of absorbing and storing heat energy than aluminum. Once the material achieves equilibrium, heat stops moving.
    Thats true, but please remember that the Varapower lights only have a battery runtime of 20-24 minutes in the highest setting. Long before that they get too hot to hold. The thermal transfer at the "neck" of the light is definitely working to some degree. If you let the light run for 1min while sitting on a table without air flow you could cook eggs on it . The one thing that these lights are missing are big heat fins to get rid of all the heat...
    When a light has a short runtime like this the absorption and storage of heat is more important than getting rid of it...

    If you ever do tests please also include heat pipes. Saabluster has mentioned that these are better than massive heatsinks.
    Last edited by The_Driver; 08-26-2018 at 03:46 PM.

  14. #14

    Default Re: Heat Management Issues

    Quote Originally Posted by The_Driver View Post
    You write a loooooot (which is good though )
    Oh yeah. None of this is easy to explain briefly. I'm trying to write out my thought process. I've always found that knowing how someone came up with the answer is more valuable than just getting the answer. I'm trying to understand the application of known principles to this specific application. There are alot of holes in my knowledge of the application. Hopeful others can help me fill in those blanks.

    Quote Originally Posted by The_Driver View Post
    What do you mean with heatsinking of the electrical contacts?
    It is common belief in this forum that soldering an emitter directly to the heatsink is better than any commonly used pcb-mounting-method. You only have one thermal barrier when it is directly soldered to the heatsink. A pcb adds at last one more to this.
    I was looking just a bit ago at an SST-90 build where the diode was directly bonded to a copper heat sink. The builder milled two slots in the top of the sink, leaving a flat surface the same width as the center block on the die bottom. He used the milled channels to route wires to the underside of the die to be soldered directly to the pads left and right of that center block. Interesting idea. But what he did was remove the thermal contact gained by having those two pads directly bonded to the heat sink. He's now got a great bond on 60 % of the diode base. Plus he's now directing heat to his supply wires. That's costing him wattage and output on top of his impaired thermal interface. Those wires are not only conducting heat, that heat is causing them to be less conductive, which is generating just that much more heat. And using current to make heat instead of light. If he would have used a star, the star's bond to the heat sink would have been the bottleneck. But by using just 5.4 of the 9 sq/mm contact surface, he moved his thermal bottleneck to somewhere else. His performance isn't any better, but he sure did a whole bunch more work.

    Quote Originally Posted by The_Driver View Post
    Thats true, but please remember that the Varapower lights only have a battery runtime of 20-24 minutes in the highest setting. Long before that they get too hot to hold. The thermal transfer at the "neck" of the light is definitely working to some degree. If you let the light run for 1min while sitting on a table without air flow you could cook eggs on it . The one thing that these lights are missing are big heat fins to get rid of all the heat...
    When a light has a short runtime like this the absorption and storage of heat is more important than getting rid of it...
    Knowing that reinforces my thought that all the copper did was delay the swiftly arriving inevitable by most likely a very small amount of time. I'm not saying that copper is bad. I'm saying it's expensive, more difficult to machine and in many instances being used where it's doing nothing. Or worse yet, distracting the builder from understanding the thermal path in his light. If copper were to be used, it should probably only be employed where it returns some sort of benefit. It becomes a bad situation where it's used expecting a benefit but the design change actually created a thermal issue elsewhere. Again, it comes down to decisions. If aesthetics override engineering, then you leave the massive fins off the light and live with brief runtimes. Like everything in life, perfection is only an illusion. Reality is the practice of compromise. We are assuming that there aren't other thermal bottlenecks beyond what we've discussed so far. I can't believe how many instances I have seen where anodizing is left in the thermal pathway.

    I grew up in the mid-west, around but not in farm country. I learned a term early on. "Amish Engineering" There were lots of Amish around. They built many magnificent barns. Incredible structures that are in many cases still standing as straight as they were the day they were built- more than a century ago. Not a one had any formal education in engineering. None had the equivalent of high school. But they worked along side of fathers and uncles and grandfathers and saw a whole lot of "what has always worked". Obviously, "what works", if you've ever been in an Amish-built barn is an incredibly overbuilt structure. This just doesn't occur with the Amish. I learned damn near everything I know from watching, looking, asking questions and looking for knowledge and information. Sometimes, people like me make assumptions without certainties behind them. Assumptions such as using materials and techniques based on seeing it done before but not really understanding completely all the considerations that went into the decision of the designer. Or the worst two words in the English language- close enough. When it's fairly low-tech, like barn structures, it's mature technology. When you're talking about taking the most advanced LED's ever made, a device that is seeing significant technological improvements in months instead of years, and then attempting to push these devices to their maximum performance, doing it successfully is going to require either alot of expensive and frustrating failures while seeking to stumble across answers, or investing the time to understand some well-established principles. I just read a thread where the builder stated he'd been reflowing diodes with lead solder since the lowest conductivity number on the chart must be the best, right? [/quote]

    Quote Originally Posted by The_Driver View Post
    If you ever do tests please also include heat pipes. Saabluster has mentioned that these are better than massive heatsinks.
    Heat pipes are interesting devices. But they don't circumvent the laws of physics, nor are they easily integrated into an elegant design. If you want a flashlight that looks like a flashlight, and have it be a reasonably portable device, there are significant challenges. If what you end up with is something that looks like a lunchbox, it's no longer a flashlight. It's a lantern. That kinda' takes us to the next step in the thermodynamic chain. All we've been talking about is conductance. getting that heat out of the flashlight requires efficiently converting all that heat energy into emittance.

  15. #15
    Flashaholic Megatrowned's Avatar
    Join Date
    Jun 2012
    Location
    Wisconsin
    Posts
    439

    Default Re: Heat Management Issues

    I love reading things like this! It makes me want to make all my lights operate "better". Unfortunately I don't have the tools or the skills to implement the knowledge I've just acquired

    So let me pose what may, or may not, be a simple question. With this 'heat management' info in mind, what would be the "best" way to make a p60 drop in fit well, so as to transfer heat well?

    What about an aleph pill (or any threaded pill, such as on a Mac's custom)? Can anything with this be improved?

    Thanks in advance,
    Austin
    Welcome to the Dark Side! Which just also happens to be the Light Side.

  16. #16

    Default Re: Heat Management Issues

    Is there any small electric fans that could pull air through the heatsink, maybe some air channels in the side of the body for it to pull air in then more air channels on other side of the fan to channel the air out. You might have to make the light a hair longer.

  17. #17
    Flashaholic* lucca brassi's Avatar
    Join Date
    Feb 2008
    Location
    Kocevje , Slovenia, Europe
    Posts
    879

    Default Re: Heat Management Issues

    the simplest way is computer fanned heatsinks - special like Midfield chip ( old northbridge or southbridge ) or special on old graphic card ( new coolers are all with big heatpipes and over 92mm fans )

    http://www.bjorn3d.com/Material/Arti...ges/TI4200.jpg

    http://www.newegg.com/Product/Produc...82E16835119080

    or just made thermal difference with some small peltier ( but comsumption will rise up ;-))

    http://si.farnell.com/multicomp/mcs-...led/dp/2253368 led on cold side and heatsinks on hotside

    -------------------------------

    or just put torch in the water like we do ;-))

  18. #18

    Default Re: Heat Management Issues

    Quote Originally Posted by CTS View Post
    But by using just 5.4 of the 9 sq/mm contact surface, he moved his thermal bottleneck to somewhere else. His performance isn't any better, but he sure did a whole bunch more work.

    .
    there are few stars made that way, copper rasied pad has heapad of the led sitting on it and contact point aren't involved much in heat transfer. it had been tested, and looked at by thermalvision cameras, and , not suprisingly, it proves beyond reasonable doubt, it does have better heat transfering abuility than convetional stars. and anyone that actually build them and experemented with direct termal contact of the heatpad will tell you , you are dead wrong. theories and reality not always go hand in hand.

  19. #19
    Flashaholic* uk_caver's Avatar
    Join Date
    Feb 2007
    Location
    Central UK
    Posts
    1,382

    Default Re: Heat Management Issues

    It would certainly seem like something that should be fairly easy to check out in practice, whether with heat measuring equipment, or by looking at the power/output of sets of LEDs mounted in various ways.

    If the people doing their best to get maximum output find they consistently get it with one mounting method compared to the others, then it would seem like that method was likely to be the best one, thermally speaking, of the ones they had tried.
    Though obviously if someone was suggesting a method which hadn't been tried, it might not be possible to rule that method out unless it was obviously worse than one which had been tried.

    Personally, I'm not sure 'bottleneck' is necessarily the best word to use - while it has some value in identifying a place it may be worth focussing some effort on, it's not a perfect description for a point in a chain of thermal resistances and might tend to confuse some people who take the word too literally.
    It's not as if heat is inexorably backing up behind a bottleneck like a moving crowd unable to get through a small doorway at more than some absolute maximum rate, just that there are some places where the thermal gradient is higher than others in equilibrium conditions.
    Even if it is not practically possible to do much about a 'thermal bottleneck' it may still be worth doing things elsewhere in the chain.

  20. #20
    Flashaholic* lucca brassi's Avatar
    Join Date
    Feb 2008
    Location
    Kocevje , Slovenia, Europe
    Posts
    879

    Default Re: Heat Management Issues

    http://web.mit.edu/lienhard/www/ahttv131.pdf

    thermal transfer book from school times

  21. #21

    Default Re: Heat Management Issues

    Quote Originally Posted by uk_caver View Post

    Personally, I'm not sure 'bottleneck' is necessarily the best word to use - while it has some value in identifying a place it may be worth focussing some effort on, it's not a perfect description for a point in a chain of thermal resistances and might tend to confuse some people who take the word too literally.

    It's not as if heat is inexorably backing up behind a bottleneck like a moving crowd unable to get through a small doorway at more than some absolute maximum rate, just that there are some places where the thermal gradient is higher than others in equilibrium conditions.
    Even if it is not practically possible to do much about a 'thermal bottleneck' it may still be worth doing things elsewhere in the chain.
    The best simplified analogy for heat is water. I will walk through a very simplified example.

    Most understand what a calorie is- the amount of energy needed to increase 1ml of water 1 degree c. Or, a BTU.

    In our example, let's say that our LED has a nominal operating temperature of 50c. In normal operating conditions the diode transfers 10 cal/min. The interface between the diode and mount (MCPCB) is capable of conducting 10 cal/min. The interface between the board and the heat sink is capable of conducting 10cal/min. The interface between the heat sink and the emissive interface is capable of 10 cal/min. The emissive interface is capable of radiating 10 cal/min. and is doing do in an environment capable of absorbing that 10 cal/min In this scenario, the diode will run at 50c until it runs out of power or lifespan.

    In comparison, we have a water pump that's output is 10 gallons a minute connected to a water faucet that is flowing 10 gallons a minute at 40 psi. into a pipe that flows 10 gallons a minute, around an elbow that flows 10 gallons a minute and into a sprayer that emits the water at 10 gallons a minute. If I were to add into that plumbing a fitting that flowed 5 gallons per minute, no matter what I have before that fitting in my flow, what's downstream of that fitting will not flow more than 5 gallons a minute. With that restriction, the pump will continue to output while the flow is restricted. Either the pressure will build until something fails or the pump will fail due to overload. If we were to place a pressure regulating vessel in the line right after the pump, the pressure as well as the workload on the pump will be maintained- until the vessel is full (saturated).

    Moving back to our LED, let's say that we place a material between the heat sink and the emissive device that flows only 5 cal/min. What will happen is with the LED continuing to receive the same current it outputs the same heat. It flows through our thermal path right up to the point that flows 5 cal/min. What will happen is as the heat flows into that heat sink, it's temperature begins to rise. The mass of that sink is capable of storing heat, just as the water pressure vessel does. It will continue to absorb heat until it becomes saturated. Somewhere before our aluminum/copper heat sink goes from solid to liquid state, the LED will have burned out, barring any other changes.

    Think of heat in terms of a unit. The analogy is a drop of water. Only so many drops of water can flow through a given restriction. This is the same for the electromagnetic energy that is heat. Obviously the molecules aren't literally flowing through the mass present in the thermal path, but if you look at the energy as an individual component a drop of heat, so to say) such as a water droplet, it does help to illustrate how heat moves within a mass. In that sense, bottleneck is probably the best way to visualize the restrictions on the conductance of heat through a thermal path.

    Using that, it's easy to find your thermal transfer issues in any design. Understanding the rates that each material conducts heat and how much area is present, it gives you a way to run some simple numbers and identify where the restrictions in your thermal path exist (it's almost always going to be at the diode base)

    Heat=water pressure. Thermal conductivity=the diameter of the pipe.
    (not considering the thermal conductance properties of each individual material)

  22. #22
    Flashaholic* uk_caver's Avatar
    Join Date
    Feb 2007
    Location
    Central UK
    Posts
    1,382

    Default Re: Heat Management Issues

    I'd have thought voltages and resistances were a simpler (and maybe more accurate) analogy.

    Assuming we're looking for an equilibrium solution for a design which can run indefinitely, if I'm looking to have a certain current (amount of heat) flowing from A to B, and there are a chain of resistances between A and B, I need a certain voltage (temperature difference) between the source and 'ground' to get the desired amount of current (heat) to flow.

    Absent any peculiar active devices, all I have is resistances, and I don't have any absolute bottlenecks (absolute flow rate limiters), just resistances that need more or less voltage (temperature) differences to get the desired flow, and the real issue is whether the overall voltage (temperature) I need requires the source to be at too high a voltage (temperature). Up to the point where things break, there is no rate at which heat which previously flowed fine suddenly starts 'backing up'.

    Even if I have a high resistance at the start of the chain, if I lower resistances further downstream, that will improve the overall situation even if only a little, and if I'm in a situation where there's not obviously anything I can do about the worst part of the chain, that just means further efforts (if any) have to go elsewhere unless it's clear the relative situation is such that any improvements wouldn't be worth the effort.

    Talking about 'bottlenecks' could mislead some people into thinking that that's where all the effort should go, and that effort anywhere else is futile.

    Regarding 10cal/min and 5cal/min, surely cal/min is the same type of unit as a Watt, and thermal paths are rated in things like Watts/Kelvin, not just Watts - within the physical limits of the materials, double the temperature difference and you get essentially double the energy flow?
    Last edited by uk_caver; 04-03-2013 at 12:06 PM.

  23. #23

    Default Re: Heat Management Issues

    Over the years I have found that most people have a pretty good grasp of how water flows through plumbing. It's a physical process they can watch and easily visualize. I have dealt with a pretty fair number of bright, trained technical people with high levels of mechanical skill that screech to a halt when they get to electrical. When walking through electrical issues with these people, I have found the water/plumbing analogy very helpful.

    Unlike electricity, heat flows through metals at a set rate. Increase the temperature- it conducts the same. In non-metallics, like a ceramic, the higher the temp, the less the conductivity.


    Another issue is that heat is stored in all matter. If I have a circuit of 12ga wire and in the circuit at some point I have a foot long, inch diameter copper bar, that bar will not store electricity. If I am conducting heat into that bar via a 12 ga piece of copper wire, it's going to be able to absorb an incredible amount of thermal energy. It would be necessary to put it into a vacuum to measure as it would radiate out faster than it would be being conducted in.

    What I hope to do is to get people to think about heat and to understand it. Once you have a basic grasp, you can run the numbers on any design and find your flaws before beginning to build. This prevents frustrating and expensive mistakes. I initially became interested in writing about this when i saw some builders making heat sinks out of brass and flashlight bodies out of titanium. I saw heat sinks being placed into anodized tubes. All represent serious mistakes based on a lack of information and understanding. Like the brass thing- copper is a great conductor so logic would have it that brass, being mostly copper, would be about the same. Not at all. Copper is four times more conductive than aluminum. Brass is less conductive than aluminum. Go figure.

    Back to bottlenecks- you really do have to look at each material interface and understand what's going on there thermally. You can take two pieces of copper, grind them micron-smooth and fasten them together and they'll conduct way less energy than one piece of copper.

    We could dig deeply "why", but that leaves all but the most hardcore out in the cold. This is supposed to be a fun hobby. I doubt most on this board want to learn about molecular physics to put together a flashlight. A basic understanding of what works and what doesn't and how to tell which from which is likely going to be more helpful. As i have mentioned before, I'm not a scientist or a physicist. I'm a guy that has to make these things function in the real world, in a practical and cost effective fashion.

  24. #24

    Default Re: Heat Management Issues

    This is a practical example of what I'm talking about-

    I was looking at a high-performance build a while back. Another poster was using it as an example. The light was available with either an aluminum heat sink or a copper heat sink. Being a larger light, the copper heat sink added about $150 to the cost. We began discussing my assertion that the copper heat sink did virtually nothing- here's why.

    In taking a holistic look at the design, the thermal paths were well thought out, but the light's ability to radiate out from the body the thermal energy generated in the diode was less than the design required. Being so, the light was only able to be run for a limited time. The copper heat sink was offered to "increase run time" But it didn't. It increased the run time of each activated interval, but the light could not be operated any longer in a given extended period. The aluminum sink saturated four times more quickly than the copper. So by that measure, it would run four times longer. But with the thermal emittance issue remaining unaddressed, the light required four times as long to cool down. In this situation, the light's inability to convert conducted heat into emitted heat was the bottleneck in the design. Nothing you did anywhere before that would change it. The materials altered the thermal behavior, but added nothing to the overall thermal performance of the light.

  25. #25
    Flashaholic* uk_caver's Avatar
    Join Date
    Feb 2007
    Location
    Central UK
    Posts
    1,382

    Default Re: Heat Management Issues

    Quote Originally Posted by CTS View Post
    Over the years I have found that most people have a pretty good grasp of how water flows through plumbing. It's a physical process they can watch and easily visualize. I have dealt with a pretty fair number of bright, trained technical people with high levels of mechanical skill that screech to a halt when they get to electrical. When walking through electrical issues with these people, I have found the water/plumbing analogy very helpful.
    Personally, I think electricity is far easier, since components are already defined in terms of relevant things like resistance, whereas plumbing fixtures typically aren't.

    I also think many people have a vague idea of how water flows through plumbing, but possibly one with all manner of misconceptions attached to it that they're not aware of.

    Quote Originally Posted by CTS View Post
    Unlike electricity, heat flows through metals at a set rate. Increase the temperature- it conducts the same. In non-metallics, like a ceramic, the higher the temp, the less the conductivity.
    Increase the temperature differential, and the flow increases basically linearly, especially where differentials are rather low in the grand scheme of things, as they typically are in flashlights.
    That's the same as electricity, where the current and voltage difference across a resistance are linearly related.

    The heat conductivity of materials does vary with temperature, but not radically so - for materials like copper and aluminium the variation in conductivity over everyday temperature ranges really isn't large.

    Quote Originally Posted by CTS View Post
    Another issue is that heat is stored in all matter. If I have a circuit of 12ga wire and in the circuit at some point I have a foot long, inch diameter copper bar, that bar will not store electricity. If I am conducting heat into that bar via a 12 ga piece of copper wire, it's going to be able to absorb an incredible amount of thermal energy. It would be necessary to put it into a vacuum to measure as it would radiate out faster than it would be being conducted in.
    Heat capacity is relevant in the short term, but it's not relevant if someone is trying to work things out for longer term equilibrium conditions, which for many flashlights are approached in normal use, and which, being harder conditions, are the ones things should often be designed for.
    It's also of little relevance when it comes to 'bottlenecks' if they tend to be at the upstream end of things, where the heat capacities of the various items are minimal. For a flashlight it may only be relevant when considering large things like the body, unless thinking about extremely short bursts of high power.

    Quote Originally Posted by CTS View Post
    Back to bottlenecks- you really do have to look at each material interface and understand what's going on there thermally. You can take two pieces of copper, grind them micron-smooth and fasten them together and they'll conduct way less energy than one piece of copper.
    I'm not saying people shouldn't try and understand things.
    I'm simply saying that by talking about 'bottlenecks', people might be misled into thinking that some particular interface or item is the only important one - it's easy for people to shift from thinking about the various things hampering heat flow to thinking about the bottleneck.

    I think that's particularly likely when people are told that particular interfaces or items have some fixed maximum energy flow rate, and that if they're in a particular thermal pathway any attempt to exceed that flow rate will cause all the excess to inexorably build up until something fails, when I don't think that's the way things actually work.

    Could you give an example of something (component or thermal interface) which has a fixed maximum heat transfer rate, regardless of the temperature difference across it, like your '5 calorie/minute' one mentioned earlier, since I'm trying to imagine something like that and I'm not succeeding.

    Quote Originally Posted by CTS View Post
    We could dig deeply "why", but that leaves all but the most hardcore out in the cold. This is supposed to be a fun hobby. I doubt most on this board want to learn about molecular physics to put together a flashlight. A basic understanding of what works and what doesn't and how to tell which from which is likely going to be more helpful. As i have mentioned before, I'm not a scientist or a physicist. I'm a guy that has to make these things function in the real world, in a practical and cost effective fashion.
    I'd have thought that the idea of adding up series resistances (electrical or thermal) to get a total resistance is hardly digging deep into molecular physics.

  26. #26
    Flashaholic* uk_caver's Avatar
    Join Date
    Feb 2007
    Location
    Central UK
    Posts
    1,382

    Default Re: Heat Management Issues

    Quote Originally Posted by CTS View Post
    This is a practical example of what I'm talking about-

    I was looking at a high-performance build a while back. Another poster was using it as an example. The light was available with either an aluminum heat sink or a copper heat sink. Being a larger light, the copper heat sink added about $150 to the cost. We began discussing my assertion that the copper heat sink did virtually nothing- here's why.

    In taking a holistic look at the design, the thermal paths were well thought out, but the light's ability to radiate out from the body the thermal energy generated in the diode was less than the design required. Being so, the light was only able to be run for a limited time. The copper heat sink was offered to "increase run time" But it didn't. It increased the run time of each activated interval, but the light could not be operated any longer in a given extended period. The aluminum sink saturated four times more quickly than the copper. So by that measure, it would run four times longer. But with the thermal emittance issue remaining unaddressed, the light required four times as long to cool down. In this situation, the light's inability to convert conducted heat into emitted heat was the bottleneck in the design. Nothing you did anywhere before that would change it. The materials altered the thermal behavior, but added nothing to the overall thermal performance of the light.
    Certainly, there's no point changing stuff just because one material is thought to be 'universally better'

    I assume that by 'radiating' you're including convection and radiation.

    When you talk about 'heat sink', are you are talking about an actual heatsink (as in fins sticking out into the air), or some chunk of metal inside the light acting as more of a heat buffer?

    And as for the comment about the ultimate heat loss to the environment being the bottleneck in the design and nothing else being able to change that, that's not necessarily perfectly correct, in that, keeping heat capacities the same, if any of the upstream thermal resistances were lowered and heat was transferred to the casing more effectively, the case would heat up faster and the LED would stay below whatever the maximum chosen temperature was for longer, even if maybe not for much longer, or even for long enough to make the change worthwhile.
    If the thermal resistance of case to environment is far higher than all the other thermal resistances put together, then changing the other ones will make little difference.
    But in order to know that you need to know what the resistances are.
    You could know from experience that the case design was inadequate for a claimed steady output, but improvements (or deimprovements) inside the case could still have an effect on the time before max LED temperature was reached.

    As for the cooling down, surely if there was good thermal conduction and significant heat storage in the lights (the heat sink heat capacity was significant relative to the rest of the light's heat capacity), if the lights were turned off at the same time, the one with the aluminium 'heat sink' would get to a lower temperature faster, but the one with the copper 'heat sink' would soon start to actually lose heat energy at a faster rate than the aluminium one because it was staying hotter, and the hotter it was the faster heat energy would be lost.
    If you turned the lights back on after a decent interval when they hadn't both cooled down to ambient temperatures, even if the copper-using one was hotter, it would take longer than the aluminium one to heat up to the point where the LED reached whatever had been chosen as the critical temperature.

    It might seem counterintuitive, but then a lot of reality is.

    Think about an extreme example, with plumbing as an analogy.
    You have two identical large sealed perfectly insulated boxes with a identical small external radiators circulating water from an internal tank, and one unit has a pint of water in its tank as a 'heatsink' whereas the other has ten gallons. Both have identical electrical heaters.
    You run each of them in turn with some heat input which is designed to be high enough (given the nature of the identical radiators and pumps) to heat the water to almost boiling, which obviously takes a lot longer with the large tank, and you turn them off and leave them for some fixed interval and then turn the heater back on.

    For the unit with the pint of water, the radiator rapidly cools it down with the heater off, and so its rate of heat loss drops quickly from the initial rate with near-boiling water to an increasingly low rate as the water cools (and the radiator cools).

    For the unit with the ten gallon tank of near-boiling water, even after a decent interval (when the other unit would be virtually cold) the water (and radiator) is still going to be hot, which means the radiator is been pumping out heat at a fairly constant high rate, much faster than the radiator on the other box did after starting to cool.

    When the heaters are turned on again, the unit with the pint of water inside would take less time to get back to almost-boiling than the five-gallon unit would when it was turned back on, even though the one-pint unit had started off cooler when the heat was turned back on, since both units would be effectively looking to the heater to make up for all the energy that had been lost (plus what was being lost while it was trying to heat up the water), and the one-pint unit had lost less heat energy in total (same cooling time, lower average rate of loss).
    Last edited by uk_caver; 04-03-2013 at 03:52 PM.

  27. #27

    Default Re: Heat Management Issues

    People are all going to understand things differently. What is clear to one might be baffling to another. All I can tell you is what works for me.

    The point here is this- a) there are plenty of examples of builds right here on this site that illustrate a lack of understanding of how heat moves, and b) I believe there's some value in getting people interested in managing the heat conductance in their projects will help them to build better lights with more ideal materials, at less cost or with less effort, or some combination of those. The more complex we make it, the fewer are going to be interested. Then what we have is a discussion between a small group that already know about the topic.

    In these designs, the Achilles's heel is always going to be the junction where the LED fastens to the heat sink. This is where you have the smallest surface area thus the point where you're trying to move the heat through the very small contact point. And even though the copper base conducts at about 400 W/mK, it's doing this through an area between 5x5mm and 10x10mm. Plus, it's now conducting into a material that's that same 5x5mm but conducts at a fraction of what the copper does. If you soldered the LED with 50/50, you're conducting at about 20 W/mK. About the best you'll achieve is 70. If you bond right to a non-insulated sink, you're only bonding about 60% of that base, so you give 40% away. If you go to a MCPCB you get full contact area but you have the loss brought about by trying to conduct heat through a non-metallic dielectric barrier. That is, unless you use a padded MCPCB. . Keep in mind that we're not talking about off-the-sheet builds. These are pushing devices for max output. Back to the LED's footprint- Here's a good thread for you to look at-- http://www.candlepowerforums.com/vb/...rmal-compounds

    This is pretty absolute- the battery discharges current, the LED converts it into light and heat. Most of the heat is conducted away from the diode while a small amount is emitted heat. Unless you reduce the current, the same amount of heat per whatever interval you want to measure will continue to flow. It has to go somewhere. It either will conduct, radiate or saturate the material it is connected to. It will impart heat into whatever it is physically connected to until that material vaporizes and is no longer capable of direct conductance. In the real world, something will fail before that.

    Back to thermal conductivity- and I'll stick to metals since that's likely all this application will be dealing with- every metal has a thermal conductance capability. It's expressed in various ways, but all of them are expressed in quantity of thermal energy, the amount of time that energy is quantified within and the physical area where the conductance can take place. Heat moves that fast. Not faster or slower or differently in different situations. X amount of energy will move at a speed of Y per square Z. Most commonly Watts per square meter in degrees Kelvin- W/mK. If something in that path conducts less and the heat continues to be generated at a set rate beyond the thermal conductivity capability equivalency of your device's thermal output, whatever mass is before this will continue to store that thermal energy.

    For your example, look at a common LED star board. Some are glass/epoxy. Some are metal since those conduct heat better. Some have thermal vias that conduct even more whereas some have an elevated center pad. Many are aluminum but some are copper, which conducts much faster. These all exist because the higher performance the diode is, the greater the need to conduct heat away from it and some do this job better than others. If you fail to use one of the better one's and push your diode, it's going to fail as a result of thermal damage. There's no way around this.

  28. #28

    Default Re: Heat Management Issues

    Quote Originally Posted by uk_caver View Post
    Certainly, there's no point changing stuff just because one material is thought to be 'universally better'

    I don't think I said that. If your light aren't overheating, why bother thinking about heat issues?

    I assume that by 'radiating' you're including convection and radiation.

    Again, unnecessarily complicating things. Convection is present, but at a much lower rate than radiation. Like I said, we can debate esoterics and absolutes and no one will benefit from this


    When you talk about 'heat sink', are you are talking about an actual heatsink (as in fins sticking out into the air), or some chunk of metal inside the light acting as more of a heat buffer?

    In the build I was referencing it had an internal heat sink

    And as for the comment about the ultimate heat loss to the environment being the bottleneck in the design and nothing else being able to change that, that's not necessarily perfectly correct, in that, keeping heat capacities the same, if any of the upstream thermal resistances were lowered and heat was transferred to the casing more effectively, the case would heat up faster and the LED would stay below whatever the maximum chosen temperature was for longer, even if maybe not for much longer, or even for long enough to make the change worthwhile.
    If the thermal resistance of case to environment is far higher than all the other thermal resistances put together, then changing the other ones will make little difference.

    Again, we can do a bunch of what-ifs. What I was trying to illustrate is that if the entire system were able to conduct heat adequately that the light would operate until the battery died, but one point in that thermal path lacked that capability, The thermal conductive ability of the entire device becomes that of the least conductive point.
    But in order to know that you need to know what the resistances are.
    You could know from experience that the case design was inadequate for a claimed steady output, but improvements (or deimprovements) inside the case could still have an effect on the time before max LED temperature was reached.

    As for the cooling down, surely if there was good thermal conduction and significant heat storage in the lights (the heat sink heat capacity was significant relative to the rest of the light's heat capacity), if the lights were turned off at the same time, the one with the aluminium 'heat sink' would get to a lower temperature faster, but the one with the copper 'heat sink' would soon start to actually lose heat energy at a faster rate than the aluminium one because it was staying hotter, and the hotter it was the faster heat energy would be lost.
    If you turned the lights back on after a decent interval when they hadn't both cooled down to ambient temperatures, even if the copper-using one was hotter, it would take longer than the aluminium one to heat up to the point where the LED reached whatever had been chosen as the critical temperature.

    Look at my example again. If you have the same amount of electrical input you're going to get the same amount of light output and thermal energy regardless of what sink you use. The choke point in the design is the ability of the device to expel heat away from the device. In that respect, the aluminum sink will absorb less than the copper one. But, the ability of the device to radiate heat out is the same, so if the copper held 3 times the heat energy, it's going to take three times as long to expel it.

    It might seem counterintuitive, but then a lot of reality is.

    Think about an extreme example, with plumbing as an analogy.
    You have two identical large sealed perfectly insulated boxes with a identical small external radiators circulating water from an internal tank, and one unit has a pint of water in its tank as a 'heatsink' whereas the other has ten gallons. Both have identical electrical heaters.
    You run each of them in turn with some heat input which is designed to be high enough (given the nature of the identical radiators and pumps) to heat the water to almost boiling, which obviously takes a lot longer with the large tank, and you turn them off and leave them for some fixed interval and then turn the heater back on.

    For the unit with the pint of water, the radiator rapidly cools it down with the heater off, and so its rate of heat loss drops quickly from the initial rate with near-boiling water to an increasingly low rate as the water cools (and the radiator cools).

    For the unit with the ten gallon tank of near-boiling water, even after a decent interval (when the other unit would be virtually cold) the water (and radiator) is still going to be hot, which means the radiator is been pumping out heat at a fairly constant high rate, much faster than the radiator on the other box did after starting to cool.

    When the heaters are turned on again, the unit with the pint of water inside would take less time to get back to almost-boiling than the five-gallon unit would when it was turned back on, even though the one-pint unit had started off cooler when the heat was turned back on, since both units would be effectively looking to the heater to make up for all the energy that had been lost (plus what was being lost while it was trying to heat up the water), and the one-pint unit had lost less heat energy in total (same cooling time, lower average rate of loss).

    See above. Heat is absolute. Thermal flow is absolute. Example- One of your devices has the ability to generate x amount of heat and to store 100X amount of heat while the emissive ability to expel .5X amount of heat, over the course of any given point of time, you will have the same energy output from a device with the same X amount of thermal generation but only 50X storage capability and the same .5X emissive ability. The first will run twice as long per interval, but the second will run twice as many intervals. In the end, whether an hour, a month or a decade, each device will have generated and expelled an identical amount of thermal energy. Just differently.
    I wrote this before- I stared writing about this to stimulate some thought about how heat works in these devices. I couldn't care less about showing anyone how "smart" I am or impressing anyone. I'm an uneducated bum that does this stuff to feed his kids. It's all hands-on knowledge.

    From here on out, I'm passing on all the theory discussion. I'm happy to talk about specifics in a particular design or about the abilities of different materials. My intent is to help people build better flashlights, not intellectual discourse.

  29. #29
    Flashaholic*
    Join Date
    Jul 2010
    Location
    Sydney, Australia
    Posts
    995

    Default Re: Heat Management Issues

    Quote Originally Posted by CTS View Post
    . . . . Back to thermal conductivity- and I'll stick to metals since that's likely all this application will be dealing with- every metal has a thermal conductance capability. It's expressed in various ways, but all of them are expressed in quantity of thermal energy, the amount of time that energy is quantified within and the physical area where the conductance can take place. Heat moves that fast. Not faster or slower or differently in different situations. X amount of energy will move at a speed of Y per square Z. Most commonly Watts per square meter in degrees Kelvin- W/mK. . . . .
    Where does W/mk involve any time factor ?

  30. #30
    Flashaholic* uk_caver's Avatar
    Join Date
    Feb 2007
    Location
    Central UK
    Posts
    1,382

    Default Re: Heat Management Issues

    Quote Originally Posted by CTS View Post
    In these designs, the Achilles's heel is always going to be the junction where the LED fastens to the heat sink. This is where you have the smallest surface area thus the point where you're trying to move the heat through the very small contact point. And even though the copper base conducts at about 400 W/mK, it's doing this through an area between 5x5mm and 10x10mm. Plus, it's now conducting into a material that's that same 5x5mm but conducts at a fraction of what the copper does. If you soldered the LED with 50/50, you're conducting at about 20 W/mK. About the best you'll achieve is 70. If you bond right to a non-insulated sink, you're only bonding about 60% of that base, so you give 40% away. If you go to a MCPCB you get full contact area but you have the loss brought about by trying to conduct heat through a non-metallic dielectric barrier. That is, unless you use a padded MCPCB. . Keep in mind that we're not talking about off-the-sheet builds. These are pushing devices for max output. Back to the LED's footprint- Here's a good thread for you to look at-- http://www.candlepowerforums.com/vb/showthread.php?348814-Thermal-compounds
    This is exactly what I mean regarding talking about 'the bottleneck'.
    A few posts ago you were talking about a light where the bottleneck was the body/environment interface, where nothing done inside the light would make any difference at all.

    And as far as the LED thermal connection is concerned, it's not just a matter of the interface materials, but their thickness.
    If you had an LED soldered directly onto a copper MCPCB, the solder will have a lower bulk thermal conductivity than the copper, but if the solder layer is just a thin film, that lower conductivity may be of limited enough importance to be essentially ignored.
    From a thermal resistance viewpoint it might take a much higher thermal gradient to push a given amount of heat through solder than through copper, but given a negligible distance of travel, that gradient might not necessitate a significant temperature difference between one side of the solder and the other, and as long as there aren't extra thermal resistances at the copper/solder boundaries, it might be that the solder could be basically ignored.

    As for a copper MCPCB, if there was a 5x5mm pillar, for each mm the pillar is high, if I calculate correctly, it has a thermal resistance of about 0.1C/W.
    Losing half of the area for direct soldering to an XM-series LED that'd be 0.2C/W per mm height, but only for a mm or so before starting to reach the main body of the MCPCB, assuming it's designed properly.

    Quote Originally Posted by CTS View Post
    This is pretty absolute- the battery discharges current, the LED converts it into light and heat. Most of the heat is conducted away from the diode while a small amount is emitted heat. Unless you reduce the current, the same amount of heat per whatever interval you want to measure will continue to flow. It has to go somewhere. It either will conduct, radiate or saturate the material it is connected to.
    I'm interested in your use of 'saturate'?
    That's not really how I'd see it - if you want a plumbing analogy, albeit not a perfect one, heat capacity is not like a tank that fills up, more like an elastic bag, which you could say is always full, but variable in size depending on the pressure.
    Electrically, it's like a capacitor - a capacitor is not 'full' or 'empty', but simply has some charge in it and voltage across it at any particular time, corresponding to the temperature of an object with a heat capacity.

    Quote Originally Posted by CTS View Post
    Back to thermal conductivity- and I'll stick to metals since that's likely all this application will be dealing with- every metal has a thermal conductance capability. It's expressed in various ways, but all of them are expressed in quantity of thermal energy, the amount of time that energy is quantified within and the physical area where the conductance can take place. Heat moves that fast. Not faster or slower or differently in different situations. X amount of energy will move at a speed of Y per square Z. Most commonly Watts per square meter in degrees Kelvin- W/mK. If something in that path conducts less and the heat continues to be generated at a set rate beyond the thermal conductivity capability equivalency of your device's thermal output, whatever mass is before this will continue to store that thermal energy.
    I think that that is, at best complicating things, if not actually mis-stating them.

    For the moment ignoring losses from the bar itself, if you have a bar made of varying thicknesses of various kinds of metals, neatly welded to each other, with an electrical heater on one end providing 10 Watts and the other end effectively perfectly cooled to some fixed temperature, the temperature of the hot end will increase until the flow of heat through the bar is 10W, at all points along the bar.

    If the input power is turned up to 20W, the hot end will eventually stabilise at a temperature where the flow rate along the bar is 20W, and the temperature difference between the ends (or between any two points on the bar) is essentially twice what it was in the first situation.

    If there was some claimed 'bottleneck' in the bar, heat doesn't really 'back up' behind it at higher power levels until something fails, it's just that in a steady state situation, the temperature differences between any two points on the bar increase or decrease in proportion to the overall temperature difference.

    If the overall temperature difference in the 10W case was 50C, and there was a short high-resistance section of bar where there was a 15C difference, in the 20W case the overall difference would be 100C and across the high resistance section it would be 30C, just as the difference across some other section that was previously 5C would rise to 10C.

    'Backing up' to me implies some disproportionate effect, and that's exactly what you are claiming when you talk about absolute limits to heat flow which cause heat to inexorably build up behind them until things melt.
    But I don't think that's how things actually are, or how people should be encouraged to imagine them by someone trying to help them understand things.

    I'm not even sure 'speed' is the best idea to raise - what's important is the flow rate (cf 'current'), and that is dependent on temperature gradients.
    That's what the K in W/mK means - for a given material, the greater the temperature difference (K), the greater the flow rate (W)
    If you were using a plumbing analogy, you would say the 'speed' changed - the greater the pressure difference along a given set of piping, the faster the water will flow.

    Quote Originally Posted by CTS View Post
    For your example, look at a common LED star board. Some are glass/epoxy. Some are metal since those conduct heat better. Some have thermal vias that conduct even more whereas some have an elevated center pad. Many are aluminum but some are copper, which conducts much faster.
    These all exist because the higher performance the diode is, the greater the need to conduct heat away from it and some do this job better than others. If you fail to use one of the better one's and push your diode, it's going to fail as a result of thermal damage. There's no way around this.
    Where did I say there was a way around it?

    And could you please give an example of a component of a thermal pathway which has an absolute maximum heat flow independent of temperature (like the 5calories/min item you mentioned earlier).
    I think it would help to increase my understanding if you could give an example, and maybe increase yours if you couldn't.

Page 1 of 2 12 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •