Results 1 to 20 of 20
  1. #1
    mtbr member
    Reputation:
    Join Date
    Jun 2007
    Posts
    52

    Heatsink Question

    Just wondering if a 10mm copper block is sufficient for 3 X cree, copper block is "shimmed" to fit snugly into a alu housing, there artic silver between all contact area and a screwed the MCPCB tightly onto the copper block. Thank you in advance!

    BTW, how hot can 3 X cree get when drive at 1amp?

  2. #2
    mtbr member
    Reputation: 02Slayer's Avatar
    Join Date
    Mar 2004
    Posts
    524
    I also have a heatsink question. Is it better to fit the heatsink with as little room as possible in the housing, or leave a tiny bit of room for some thermal paste/epoxy? And if so, how much room?
    I have an Al disk I am shaping to fit inside the tube of my housing and right now as soon as I take off another 1/2 millimeter it will be a tight, like I'll need to force it in, fit. My thought is that there wont be any room for thermal paste, but I'm not convinvinced the "touching" surface between the tube and the disk inside it is sufficient for thermal transfer. By strict area formula, it will be about 160 square mm surface interface. The disk will be about 33 grams. Help?
    Hope I'm not hijacking your thread. I thought our questions were kinda similar and would draw qualified responders with useful knowledge and experience.
    Wrecker of fine things

  3. #3
    Singletrack Daydreamer
    Reputation: rockymtnway's Avatar
    Join Date
    Nov 2004
    Posts
    346
    @ Rott: Not sure what you mean by a 10mm block. If it's truly a 10mmx10mmx10mm cube, then that's likely enough for one LED (with free airflow around it) but not for three. If you're talking about a 10mm thick disc in a 35mm housing with good contact with the aluminum with a thin coating of thermal grease or epoxy, it's probably is adequate.

    As for how hot a Cree can get at an amp, I learned quickly that if you're holding a Cree on a board in your hand at 700ma, you'll have a blister on your finger in under 15 seconds. Don't try this one at home kids.

    @ Slayer: You want direct contact. Heck, ideally, you want one piece of metal. Copper has a conductivity of 401, Aluminum is 237, thermal grease is 0.7-3.0. With that in mind, you may wonder why we use thermal compound at all, but air is one of the best insulators on earth, with a conductivity of 0.025. Try to have perfect contact and use just enough thermal compound to fill any tool marks.

    Look at the section in Wikipedia on thermal conductivity and follow some of the related links for some good info (albiet, much of it is over my head).
    Train 'til you puke. Cheat to win. Party like a rockstar. We miss you, Jan!

  4. #4
    mtbr member
    Reputation:
    Join Date
    Mar 2007
    Posts
    120
    10mm copper block mightl be a bit overkill and heavy -5mm aluminium would be sufficient.
    Of course if you don't mind the weight -the copper will be great.
    The more important thing is surface area to suck the heat quickly away from the heatsink and into the air.


    Any thermal transfer compound is a good idea -unless you have a 100% press fit ,the compound will fill any gaps. The less room between the mating metal surfaces -the better.

    Cheers
    Dom

  5. #5
    mtbr member
    Reputation: 02Slayer's Avatar
    Join Date
    Mar 2004
    Posts
    524
    Quote Originally Posted by rockymtnway

    @ Slayer: You want direct contact. Heck, ideally, you want one piece of metal. Copper has a conductivity of 401, Aluminum is 237, thermal grease is 0.7-3.0. With that in mind, you may wonder why we use thermal compound at all, but air is one of the best insulators on earth, with a conductivity of 0.025. Try to have perfect contact and use just enough thermal compound to fill any tool marks.

    Look at the section in Wikipedia on thermal conductivity and follow some of the related links for some good info (albiet, much of it is over my head).
    Thanks. While the work is slow right now, I'm trying to stay patient and do things as best and smart as I can. I'm just worried that the curved surfaces of the disk and the cylinder won't sit nearly as flush as two flat planar surfaces, and while the disk and cylinder appear to be in contact, the actual surface area of contact will be tough to accurately determine and/or rely upon alone.
    Maybe Achesalot square design was the way to go.
    Wrecker of fine things

  6. #6
    Singletrack Daydreamer
    Reputation: rockymtnway's Avatar
    Join Date
    Nov 2004
    Posts
    346
    Agreed that Achesalot design is probably the best route for anyone without access to a CNC machine, but I still plan on proceeding with a MR-11 conversion project (in addition to some Achesalot variants). One option for getting around piece of material (although not necessarily the perfect size) is using a hole saw. Sure you have a hole in the center from the pilot hole, but otherwise, it starts you with something very round that you can trim rather than just banging off corners on something square.
    Train 'til you puke. Cheat to win. Party like a rockstar. We miss you, Jan!

  7. #7
    mtbr member
    Reputation:
    Join Date
    Oct 2007
    Posts
    66
    Finally, a topic on the forum I can speak to intelligently, well somewhat anyway.

    I'm a noobie to MTB but I have years of experience with heat transfer and thermal management of electronics. As for the specifics of the application being discussed, well, I'm at a loss. So if you could elaborate or point me to pics of what you're trying to do I can probably be a lot more help. I don't know what a "Cree" is, nor do I have any clue what an "Achesalot design" is.

    As for Cu vs. Al, well Al should suffice for most applications unless you have very high heat flux and need to spread the heat extremely rapidly, usually required for components with lower operating temperature limits.

    As was mentioned, the most significant affect will be surface are exposed to airflow and this is the choke point in air cooled designs. Thermal capacitance of air is low and the effective convective heat transfer coefficients relatively low but depends on air velocity flowing over the surface. So, rather than solid blocks, finned heat sinks are much more effective. A pin fin heat sink is best if the air flow direction varies.

    Now, on the topic of thermal grease, yes use it. No matter how close a fit the two parts have or how smooth the surface finish, microscopic roughness exists and creates interstitial pockets of air, a poor thermal conductor. The thermal grease fills the pockets and improves the heat transfer across the boundary. You want as tight a fit as possible between the parts, DO NOT leave a gap for the grease, let it squish out. The more direct contact between the material the better. Let the grease fill the small gaps.

    Sorry if I rambled too long. Hope the info helps.

  8. #8
    mtbr member
    Reputation: heatstroke's Avatar
    Join Date
    Jul 2003
    Posts
    1,007
    Swamp Rat.
    Don't know if you can help here.
    I have a cutters mr11 3xQ5 light. I've got a post somewhere on Mtbr.

    The leds fixed to a 7 mm thick, 35mm diameter Al plate (lapped the mcpcb to the plate and then applied thermal grease + 3mm bolts)

    The plate is the fixed to a finned 50mm long (compression fit). Fins are 4mm deepx2mm widex2mm spacing alone 1/2 the can.

    Are there any calculators out there that can tell me how much heat this can dissapate, i.e the Delta T between plate and ambient ?
    Any rules of thumb, I do not have access to a decent thermometer which can measure the temp near the leds.

  9. #9
    mtbr member
    Reputation:
    Join Date
    Oct 2007
    Posts
    66
    heatstroke, let me noodle on this. This is the post you were referring to? Yes there are ways to determine the LED temp analytically but the problem is understanding the heat transfer characteristics from the exterior of the body to air. This is entirely dependent on the dT between them and the velocity of the air flowing over it, and direction for that matter, as well as the density of the air. Convection is much more difficult than conduction. The entire thing is a series network of thermal resistances that can be added together. You then get a deg/W value of temperature rise, if you know the power and you know the ambient you simply do:

    Tf=Ta+Rt*P
    Where: Tf= Final temp (LEDs) Ta=ambient : Rt = total thermal resistance : P = Power

    Now the trick is to know each Rx (thermal resistance in the system). For metals and interfaces it's fairly simple. It's the final Rsa, thermal resistance of the sink to air that get difficult. That has to be determined empirically w/ measurements, or analytically with a CFD (computational fluid dynamics) software tool.

    Oh, and to answer your question, "how much heat can this dissipate" several kilowatts
    Of coarse I'm being facetious, I get asked this question at work all the time and it is never properly framed, not to mention straight forward. At some point equilibrium will be established and the heat will dissipate. The real question is, "how much heat can this dissipate and maintain the LEDs within their specified operating range".

    One other note on your design, which is very nice work by the way. I would recommend making the fins along the axis of the housing rather than around the circumference. This way the air will flow down the length of the fins and dissipate the heat much more efficiently. I assume you are riding in the direction the light is shinning and not sideways

    I'll try and dig up info on the specifics of the MCPCB to understand that thermal path. If you can direct me to a pic that would help. I need to know how the heat gets off this PCB assembly and into the 7mm thick heat sink on the back end.

    Sorry, another long post and probably more info than you really wanted.

    Oh, and I do have access to a CFD tool, so if I can get enough info I can model this, it is very simple assembly, and show you some really cool images of what's going on w/ the heat in the system. I've been looking into lighting and it's all very expensive, so if this can be a lower cost approach I have a vested interest in the effort.

  10. #10
    mtbr member
    Reputation: heatstroke's Avatar
    Join Date
    Jul 2003
    Posts
    1,007
    I've been trying to locate the info on the MCPCB . Here is a link from cutters:
    http://www.cutter.com.au/products.php?cat=41
    http://www.cutter.com.au/proddetail.php?prod=cut759
    I am going through this link to try and find out more :
    http://www.candlepowerforums.com/vb/...d.php?t=140996

    The MCPCB is 1.6mm thick, 34 mm in diameter. Right now I do not know anymore than that ;-(

    I wish I had a mini mill or a mini mill attachment to the lathe , then we could create length wise fins instead !
    Last edited by heatstroke; 11-21-2007 at 09:52 PM.

  11. #11
    mtbr member
    Reputation:
    Join Date
    Aug 2007
    Posts
    33
    I've had my bflex light going for about 3-4 weeks now and experienced not heat problems. My light case is a 4" long. The handle is about 1x2.5" and the head diameter is about 1.7" diameter. No Cooling fins. You can calculate the area if you want. I have checked the light during and after 1.5 hr rides and the temperature is hot but not uncomfortable to the touch. I'm not giving advice or even a counter point. Just sharing my experience with my light. In my industry the basic formula for heat flow is A x K x T1-T2 = BTU's/min A=area, K is the conductivity T1-T2 the delta Temp. I had no desire to "engineer" the light - I just did some testing. A few more thoughts;

    1) My light dissipates heat well cause I used thermally conductive potting epoxy to get the heat from the heat sink to the case. And I don't stop for long periods on high beam.

    2) I've been riding on nights at 60-65Deg F. On a hot summer night -say about 80 the light is going to run hotter. My gut feeling at this point is thats not going to be a problem. But don't burn your light on my gut feeling.

    3) More speculation. You going to burn your skin before the light burns up. Unless your heat sink is totally non functional.

    4) From the Cree data sheet.
    "Based on internal long-term reliability testing and standardized forecasting methods, Cree projects XLamp LEDs to maintain an average of 70% lumen maintenance after 50,000 hours, provided the LED junction temperature is maintained at or below 80°C." thats about 176F any body know the blister temp for skin? Lets say running the light 1000mA at a high temp shortens the usable life to 1/4. Thats almost a 1.5 years of straight bike riding. I think my bike will fall apart first.

    I've got a Fluke dual channel meter. and 30 Gage K thermocouples I could inbed in the epoxy. So maybe I''ll log some ride data and post it on a graph.

    I had allot fun building this light and my advice to anybody thinking about doing it is to go for it. I think someone could build this for 100.00 if you did the footwork (shopping around). Look around at the different designs that have been posted. Lots of good ideals. And if you cut some corners and it burns up - just toss it in the trash when nobody looking.

  12. #12
    mtbr member
    Reputation: heatstroke's Avatar
    Join Date
    Jul 2003
    Posts
    1,007
    I noticed that the light was an acceptable warm when riding.
    However when I fired up the light for 5 minuted stationary - the light was uncomfortably hot. I set the bflex trip at 70deg C and it had not tripped yet. Since the bflex is only "floating" behind the the heatsink plate, I don't fancy ramping it up to 1A until I have a reasonable confidence that the design is robust. Getting the CREEs to Malaysia is a long wait.

    Unfortunately (?) the light has been popular among my riding friends. So I am now machining up the 5th light at cost for them. I did not have enough bar/ and pipe stock so I had to limit it to 5 hard core riders that can't afford to splurge on some off the shelf bling. Plus I was getting bored churning the same thing out again and again.

  13. #13
    mtbr member
    Reputation:
    Join Date
    Aug 2007
    Posts
    33
    Malaysia?
    Dam I bet you have some killer trails in Malaysia.

  14. #14
    mtbr member
    Reputation: scuppy's Avatar
    Join Date
    Nov 2007
    Posts
    94
    Although Cree wont commit to heatsink recommendations, Lumelids/Luxeon will. They are essentially the same technology (Despite some myths, the Luxeon Star 5 emitts 80lm @ 350mA versus Cree XR-E emitting 80lm @ 350mA, although the Vd typ is slightly lower). The recommendation for luxeons in their thermal management doc is 1 square inch of heatsinking surface area per watt. This includes the base plate, so a 1 watt LED doesn't need extra heat sinking. Obviously this is subjective based on air movement, ambient temp etc etc, but given bicycles are usually moving, it's probably a safe bet.

  15. #15
    mtbr member
    Reputation:
    Join Date
    Oct 2007
    Posts
    170
    Something that has not been mentioned in this thread yet is that getting heat away from the led is not just about stopping it popping, its also about getting the most light from the led. I've seen data from tests that show significantly lower light levels as temperature increases. The lower you can keep the temperature the brighter the light it will emit.

    I'm actually trying to work out how thick to make the plate holding my cree's to provide effective heat transfer to the shell of my housing so really interested in this thread. Finding the delicate balance between function and weight and trying to do it without just guessing.

    Thanks for the information folks, much appriciated

  16. #16
    mtbr member
    Reputation:
    Join Date
    Aug 2007
    Posts
    33
    Good point iggs.

    This application note describes the types of failures common to high-power LEDs

    Cree® XLamp® LED Reliability
    http://www.cree.com/products/pdf/XLamp_Reliability.PDF


    Heat sink info.

    Cree® XLamp® LED
    Thermal Management

    http://www.vs-optoelectronic.com/ima...Management.pdf

    Did not spend allot of time going over this stuff. The general message I get is if the LED is kept below 85C I'll get more useful life than I need out of this thing.

    I'm going to log a 1.5 hr ride to see were the light is operating. Sometime this weekend.

  17. #17
    mtbr member
    Reputation: heatstroke's Avatar
    Join Date
    Jul 2003
    Posts
    1,007
    Sidetracking a bit. Yes the trails are good and there are lots of it (old logging trails)
    BUT because it is so heavily logged, there is naff all to see. And it is not as hilly as I'd like.
    If you hit google earth - search for Miri Malaysia and you can see all of it. Over the past year we haave been tracking all our trails and digitizing what we can see on GE. We've ridden about 95% now.

  18. #18
    mtbr member
    Reputation:
    Join Date
    Oct 2007
    Posts
    66
    stewed, you are right that you don't need to go to great lengths to get something that works. heatstroke asked so I was getting down to the details and not to mention I'd like to apply what I have to do to earn a living to some fun endeavors as well. Thanks for the links on the data & info, I'll review them. It's also interesting that this effort can help get more light from the LED as well, so it's not just a function & reliability effort.

    heatstroke, I certainly understand about the radial fins if you are using a lathe. I guess I just assumed you made it on a mill.

    As I have been reading through some of the other posts in these DIY lights, I'm learning more. I'll try and build up a CFD model in the next few days and refine it as more detailed info becomes available. You never know what we can learn from this or what ideas this may spawn from others.

  19. #19
    mtbr member
    Reputation:
    Join Date
    Jun 2007
    Posts
    52
    Malaysia is a great place to ride. I'm there once a month riding Johore. Never a problem there despite what other advise. Find the trails solid, the people heartwarming and the food superb and cheap!

    Now..did a little bench test of my own. I swop my 38mm diameter X 10mm thick copper heatsink for 38mm diameter X 6mm thick aluminium heatsink. This is what I get after an hour, siting on desk without moving air. Ambient Temperature is 29 degree celsius. Temperature is measured at the other side of the heatsink (opp MCPCB) with a thermocouple probe ripped from a Hardcano 11.

    At 350mah - 33 degree celsius
    At 500mah - 58 degree celsius
    At 750mah - 70 degree celsius in 11 minutes
    At 1000mah - 70 degree celsius in 4 minutes

    At below 500mah, temperature remain constant even after 4 hours.

    I mounted the MCPCB to a Copper CPU heatsink, running the fan at 5800rpm. At 1amp, the temperature went up to 70 degree celsius in 51 minutes.

  20. #20
    Singletrack Daydreamer
    Reputation: rockymtnway's Avatar
    Join Date
    Nov 2004
    Posts
    346
    Great data! Do you have the same data for when you were using the copper heatsink? Also, I assume this was not inside a sealed housing, correct?
    Train 'til you puke. Cheat to win. Party like a rockstar. We miss you, Jan!

Posting Permissions

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