Mounting of Leds - thermal performance- Mtbr.com
Results 1 to 15 of 15
  1. #1

    Reputation:
    Join Date
    Jun 2009
    Posts
    82

    Mounting of Leds - thermal performance

    What is the best method of mounting leds (such as Cree MCE), to minimise the thermal resistance from the solder point on the led, to the housing?

    The calculation is simple and robust, provided the thermal resistance of the materials are know, which in fact they are.

    The "industry standard" method is to solder the led to an MCPCB (metal core printed circuit board) disc or star, and then firmly clamp the rear metal surface of the MCPCB to the heatsink, which in our case is the casing. Cree and Arrow state that the thermal resistance of "typical" MCPCBs is around 0.8 to 1.0 DegC/W At the full-power MCE dissipation of about 10W, this represents a temperature difference of 8 to 10 DegC, which is workable but not desirable. The thermal resistance of the MCPCB-to-casing interface is negligible, because this is a metal-to-metal interface with very large surface area compared to the solder point of the led.

    I apologise if what I am about to write has been said before, but it strikes me as important to know how that compares to the DIY approach of gluing the led directly to the case/heatsink. The best epoxy available to my knowlwdge is Arctic Silver Adhesive filled with silver particles, with a quoted thermal conductivity of 7.5 W/mK. Note that the "AA" Arctic Alumina epoxy is not as good, and therefore should not be used IMHO. Note that, by comparison, aluminium has a thermal conductivity of about 200, and copper is 400, so even the very best thermally conductive epoxy is a VERY POOR thermal conductor compared to metals. For an MCE, the area of the solder point on the back of the led is 14 square mm. Assuming an adhesive thickness of 0.1mm, the thermal resistance of the interface will be 0.95 DegC/W, which is comparable to an MCPCB, and will result in an additional 9.5 degrees of semiconductor junction temperature at 10W dissipation. The key parameter here is the bond thickness, which is hard to control. Basically, if you can manage to get the bond thickness less than 0.1mm then thermal performance will be good, but if greater than 0.1mm then performance will be poor, unacceptably so in my view. The bond thickness will tend to be set by how much adhesive you apply, and it is very hard to know what kind of bond thickness most of the DIY people are achieving. Anyone care to make a guess?

    The best mounting method which I will probably use myself is to build your own copper mounting plate, say around 20mm square, and directly solder the led to this plate. In this case, the adhesive is in effect replaced with solder, which has a thermal conductivity of 49 W/mK. For the same 0.1mm bond thickness, the thermal resistance of the interface will be 0.15 DegC/W, corresponding to a negligible 1.5 DegC temperature differential. As previously stated, the thermal resistance from this plate to the heatsink/case will be miniscule and can be neglected. As this plate can be firmly srewed or clamped to the heatsink/casing, the gap will be very small, say 0.04mm or even less. If this small gap is filled with Arctic Silver adhesive or paste, then for a 20mm square plate, thermal resistance will be 0.013, or 0.13 DegC at 10 Watts, totally negligible. The copper plate has 2 other advantages - it can be removed or relocated, and it helps spread the heat to the outer surface of the casing, again reducing overall thermal resistance.

    I am not trying to discount the use of epoxy, but be sure to use the silver filled Arctic epoxy, not the alumina filled, and make a great effort to reduce the bond thickness as much as possible.
    Last edited by cdcdcd; 07-09-2009 at 12:05 AM.

  2. #2
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    169
    If it were only so simple...

    Aluminum alloys don't conduct as well as pure Al. 6061 is between 150 and 180 depending on the temper. 6061 is actually pretty good compared to some others that can be as low a 100. Makes you wonder what kind of Al heat sinks are made of.

    The solder is only around 50. I'm sure that is way better than any thermal epoxy, but it'd not as good as Cu or Al.

    If you solder the MC-E to a copper plate what are you going to do with the leads? They are in the same plane as the backside thermal connection. You would have to bent them up or put some lacquer on the isolate them from the copper basse. Maybe some Kapton tape. It can kind of take soldering temps Sort of like making your own MCPCB.

    Now if someone would just make a star board from copper our LEDs could run several degrees cooler.
    Last edited by vroom9; 07-08-2009 at 11:38 PM.

  3. #3

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by vroom9
    If it were only so simple...

    Aluminum alloys don't conduct as well as pure Al. 6061 is between 150 and 180 depending on the temper. 6061 is actually pretty good compared to some others that can be as low a 100. Makes you wonder what kind of Al heat sinks are made of.

    The solder is only around 50. I'm sure that is way better than any thermal epoxy, but it'd not as good as Cu or Al.

    If you solder the MC-E to a copper plate what are you going to do with the leads? They are in the same plane as the backside thermal connection. You would have to bent them up or put some lacquer on the isolate them from the copper basse. Kind of like making your own MCPCB.

    Now if someone would just make a star board from copper our LEDs could run several degrees cooler.
    Yes, I'm aware that the aluminium alloys have lower conductivities but it makes no difference to anything I said, so I believe the practical situation is as simple as I have stated.

    The copper surface will be milled to produce a raised section in the middle where the led is soldered on, so the leads will be 1mm or so above the copper surface. Then I'll solder wires directly onto the leads, exactly as the DIY people do anyway when gluing the leds directly to their aluminium housings. Not difficult, really.

    The problem with the commercially available star boards is not that they are made of aluminium per se, but because they electrically isolate the copper heatsink connection from the aluminium substrate, which is not necessary with modern leds such as the MCE which have an electrically isolated thermal path. This was explained in one of the later postings in the "measurement of die temperature" thread.

    I completetly agree with you though, that it's a shame we can't buy decent, non-electrically-isolated star boards with a copper substrate, where the led solders directly to the substrate. Am I missing something, or is this just common sense?

  4. #4
    I like bloody ankles
    Reputation: dsjc's Avatar
    Join Date
    May 2007
    Posts
    139
    Quote Originally Posted by cdcdcd
    The bond thickness will tend to be set by how much adhesive you apply, and it is very hard to know what kind of bond thickness most of the DIY people are achieving. Anyone care to make a guess?
    Probably quite a bit thicker than we'd like or realize in most case, unless you are milling your own housings and gluing the emitter direct to the milled surface.

    I've noticed that all the stars I've purchased, as well as extruded aluminum stock like square tubing or flat bar, have a concave shape to them. So unless we take great care to lap all surfaces flat, there will naturally be gaps that the epoxy must fill, and that could easily be >1mm from what I've experienced.

    Even in the case of perfectly flat surfaces, it takes quite a bit of grinding the star down into the epoxy to make the layer optimally thin - probably more time required than the 5 minute set epoxy will allow before it gets too stiff.

    Out of curiosity, how will you solder the emitter to the sink, must this be done using reflow ?

  5. #5
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    270
    Quote Originally Posted by dsjc
    Out of curiosity, how will you solder the emitter to the sink, must this be done using reflow ?
    Yeah, I am wondering also how to solder emitter to aluminium...

    BTW, looks like with sandpaper 400 I was able to achive a very flat surface and managed to make Arctic Silver layer very thin, probably less than 0.1mm.I used a spare collimator from DX to press the MC-E to the body.

  6. #6
    mtbr member
    Reputation:
    Join Date
    Feb 2009
    Posts
    169
    Quote Originally Posted by Itess
    Yeah, I am wondering also how to solder emitter to aluminium...
    It would have to be done with some kind of reflow. The hard part of soldering to aluminum is the oxide layer. It usually takes nasty fluxes. In theory one could get the solder on the Al and then clean it real well to get all the flux off. Then reflow the LED to the pre soldered Al.

    I don't think that "normal" tin/led 63/37 solder works with Al. Some more specialty solder made just for Al is needed. Anything with tin will have corrosion issues and the zinc based ones take more heat to melt.

    No matter what it is always a pain to solder to Al. In this application it would be easy to end up frying the LED in the process. I'm sure that it could be done though.

  7. #7

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by dsjc
    Probably quite a bit thicker than we'd like or realize in most case, unless you are milling your own housings and gluing the emitter direct to the milled surface.
    ?
    Yes, except that gluing the bare emitter is precisely the situation when a thin bond is critical, because the surface area is so small. My guess is that most people end up with a bond thickness that is set by the amount of epoxy that is applied, because it would be very difficult to firmly press the small, fiddly emitter down with sufficient force to squeeze out the excess. Actual bond thickness obtained - who knows ?


    I've noticed that all the stars I've purchased, as well as extruded aluminum stock like square tubing or flat bar, have a concave shape to them. So unless we take great care to lap all surfaces flat, there will naturally be gaps that the epoxy must fill, and that could easily be >1mm from what I've experienced.
    Here we are talking about the epoxy bond between the star and the housing, which fortunately is less critical re bond thickness. Even so, a 1mm gap would be undesirable, though I think this would be a very extreme case. For an MCE dissipating 10W, a 1mm gap using Arctic Silver epoxy would result in a temperature difference from 20mm star to housing of around 4 DegC - more than it should be but still not really a problem. I conclude that the interface from star to housing is not an issue unless the gap is ridiculously large such as 1mm or more.


    Out of curiosity, how will you solder the emitter to the sink, must this be done using reflow ?
    Yes, easy enough with a reflow oven or temperature controlled hotplate. What I will do, though, is make jig for locating the emitter.

    The jig will consist of a square aluminium plate, approx 15x15x2mm,with a round hole in the middle to drop over the round lens assemply on the emitter. There will be a 2mm holes in each corner of the plate, and matching holes drilled and tapped into the surface on which the emitter is to be mounted. Therefore the jig will not only accurately locate the emitter, but will provide the capability to clamp the emitter down with considerable force if desired.

    I will experiment with using Arctic Silver epoxy, because with this jig forcing the emitter onto the surface, I strongly suspect I can easily and reliably achieve bond thickness of 0.025mm or less. I will accurately measure the bond thickness achieved by measuring the total thickness of led+substrate, with and without epoxy. If epoxy bond thickness of 0.025mm or less is obtained, I won't bother with soldering or copper stars at all, and will glue the emitters directly to the housing. To be honest, I would prefer not to shag around with soldering the emitter if good results can bo obtained with epoxy.

    I am, however, slightly concerned about whether the differential expansion between the copper solder point of the emitter, and the aluminium housing, may eventually cause failure of the bond. I'll email Cree and ask their advice about that.

    Hopefully my MCEs and Arctic Silver will arrive today, so I can get moving and make useful progress over the weekend.

  8. #8

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by Itess
    Yeah, I am wondering also how to solder emitter to aluminium...

    BTW, looks like with sandpaper 400 I was able to achive a very flat surface and managed to make Arctic Silver layer very thin, probably less than 0.1mm.I used a spare collimator from DX to press the MC-E to the body.
    None of my suggestions involve soldering to aluminium, which would indeed be very difficult ....

    If I make my own "star", then it will be made from coppper.

    If I mount the emitter directly to the aluminium casing, then it will be using epoxy, and with the aid of a clamping jig as described in my previous posting.

    I'll report back with my findings.

  9. #9
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    I have been watching this thread with great interest where as the maths is my achiles heel most of my results are a lot less scientific but do seem to have good results.

    I read all the threads and use the info presented by all to good effect and I thank you all
    for doing the reshearch .


    I have been epoxieing leds direct to the housing for 2 years now with real good results
    I have made jigs for this purpose from old optics which greatly help the positioning and squeezed in a clamp the bond layer is very thin and what squeezes out does form a reinforced edge to help hold the led.

    Please correct me if I am wrong but is Arctic silver conductive so probably not one to use if it can get on the contacts , but for an mce would be fine

    some folks use the thermal paste on the led with epoxy on the edges to hold it firmly

    Copper would be great but is not a good metal to machine and for a bike light not IMHO worth the extra effort

  10. #10

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by troutie-mtb
    I have been watching this thread with great interest where as the maths is my achiles heel most of my results are a lot less scientific but do seem to have good results.

    I read all the threads and use the info presented by all to good effect and I thank you all
    for doing the reshearch .


    I have been epoxieing leds direct to the housing for 2 years now with real good results
    I have made jigs for this purpose from old optics which greatly help the positioning and squeezed in a clamp the bond layer is very thin and what squeezes out does form a reinforced edge to help hold the led.

    Please correct me if I am wrong but is Arctic silver conductive so probably not one to use if it can get on the contacts , but for an mce would be fine

    some folks use the thermal paste on the led with epoxy on the edges to hold it firmly

    Copper would be great but is not a good metal to machine and for a bike light not IMHO worth the extra effort
    Your experience is extremely valuable, and your lights are the standard by which others are judged.

    We appear to independently come to the same conclusion. If a jig/clamp is used to squeeze out the excess epoxy, forming a very thin layer of epoxy betqween led and heatsink, then this is a good method, with excellent thermal performance. I emailed Cree and will be interested to see what they say, but the bottom line is that you have been directly epoxying high-power leds for some time and have achieved thin bonds,and had no problems with epoxy bond failure.

    Arctic silver is not conductive as such and should not cause problems for bonding leds, even if touching the pins(s) of the led. That said, Arctic silver does increase capacitive coupling so could affect sensitive high impedance, high frequency circuitry. Fortunately a power led is essentially a low-impedance DC device, so Arctic silver should have no effect. The thermal conductivity of the silver filled Arctic Silver epoxy is nearly twice as good as the totally electrically insulating alumina or ceramic filled Arctic epoxies, so appears to be the best choice. Does anyone have bad experience,or know of any reason why the thermally superior silver filled expoxy should not be used to mount the led?

    My MC-E leds and square ledil CMC optics have arrived, so I have spent the day milling a rectangular block of aluminium, which will eventually become a dual MC-E bike light. I hate to admit it, but this is a fun project, and the light should be useful as well.

  11. #11

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    As promised, here is reproduced my email to Cree, and their reply :-



    Hello,

    We have some engineering questions re optimum thermal solutions for mounting Cree MC-E multichip leds.

    Firstly, the Cree literature states that the MCE has an "electrically isolated thermal path", which presumably means that the solder point is electrically isolated from anode and cathode. Is this correct?

    Answer: It is correct.

    Assuming this is true, it seems strange that the commercially available MCPCBs for the MC-E electrically isolate the copper solder point on the MCPCB from the aluminium substrate of the MCPCB. As a result of this redundant electrical isolation on the MCPCB, the thermal resistance across the MCPCB appears to be typically 0.8 to 1.0 K/W. As a result, at the maximum MC-E dissipation of around 10W, the temperature rise across the MCPCB will be around 8 to 10 DegC. While this is satisfactory in most applications, it is clearly not optimum. Are we correct in saying that a superior thermal solution would be to use a copper substrate on the MCPCB, and solder the MC-E directly to this copper substrate? This should result in the semiconductor junction temperature of nearly 8 DegC lower than it would be with existing electrically isolated MCPCBs.

    Answer: You are right. The commercial available MCPCB is designed for all kind of LEDs (include no electrically isolated LED). With MCE, you can do the way you described to take the advantage of electrically isolated thermal pad.

    To avoid the cost and thermal penalty of an MCPCB, in our application we are considering the use of a thermally conductive epoxy to bond the MC-E emitter directly to an aluminium heatsink. The thermally conductivity of the adhesive is 7.5 W/mK, and the area of the MC-E solder point is 14 mm^2, with the result that the thermal resistance of the epoxy interface will be around 0.95 K/W for a bond thickness of 0.1mm, which is comparable with an MCPCB. However, bond thicknesses of down to 0.025mm are obtainable giving lower thermal resistance than an MCPCB, at lower cost.


    Is epoxy bonding of the MC-E emitter, as described, a practical solution? In particular, are there likely to be issues of bond failure over many thermal cycles, given the difference in expansion coefficient between the aluminium heatsink and the MC-E solder point? If so, is it satisfactory to epoxy bond the MC-E directly to copper, rather than aluminium?

    Answer: This is very good question. We never tried the way you described. It will depend on the operation conditions of the LEDs. In normal condition, it should be fine.




    If I may be so bold as to summarise this thread, commercially available MCPCB stars for the MC-E are poorly designed and have unecessarily poor thermal performance, and a better thermal solution is to epoxy the MC-E directly to the aluminium heatsink/casing, using Arctic Silver epoxy, taking care to ensure that the bond thickness is as thin as possible.

    I find it interesting that apparently Cree has never tried the "direct epoxy" mounting method, so arguably the DIY people here are leading the way.

  12. #12
    mtbr member
    Reputation: yetibetty's Avatar
    Join Date
    Dec 2007
    Posts
    1,781
    I have epoxied direct to the aluminium case for years using an old optic and very light preasure from a small vice to hold it all down while the Artic Alumina cured, in fact I did it on my first ever build and that light has been droped many times and is infact now used to light my work bench (got to do somthing with the first one)and is attatched to my drill press so has put up with lots of vibration, no air flow and I often forget to turn if off for a whole day so it has got very hot and the AA is still holding strong after 3 years...and yes the heat is transferd much quicker than through an MCPCB, however Arctic silver 5 gets the case even hotter and quicker but it isn't a glue so you require constant preasure on the optic to hold things in place. I use an O ring between optic and face plate or screw on bezel wih an O ring.

    The Led's that I used then were SSC p4's and had a seperate slug with legs for the wires (like the MC-E) so it was easy to do.

    I have never done it with a Cree XRE due to the bottom also having the conections.

    Just my findings.
    Last edited by yetibetty; 07-13-2009 at 08:23 PM.

  13. #13

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by yetibetty
    I have epoxied direct to the aluminium case for years using an old optic and very light preasure from a small vice to hold it all down while the Artic Alumina cured, in fact I did it on my first ever build and that light has been droped many times and is infact now used to light my work bench (got to do somthing with the first one)and is attatched to my drill press so has put up with lots of vibration, no air flow and I often forget to turn if off for a whole day so it has got very hot and the AA is still holding strong after 3 years...and yes the heat is transferd much quicker than through an MCPCB, however Arctic silver 3 gets the case even hotter and quicker but it isn't a glue so you require constant preasure on the optic to hold things in place. I use an O ring between optic and face plate or screw on bezel wih an O ring.

    The Led's that I used then were SSC p4's and had a seperate slug with legs for the wires (like the MC-E) so it was easy to do.

    I have never done it with a Cree XRE due to the bottom also having the conections.

    Just my findings.
    Why do you use the "AA" Arctic Alumina epoxy, rather than the thermally superior silver filled Arctic Silver epoxy, which has almost twice the thermal conductivity (4.5 vs 7.5)?

    I notice people often referring to "AA", and am not sure if they are just throwing around a buzz term, when what they really mean is Artic Silver Epoxy?

    Why would anyone use Arctic Alumina epoxy?

  14. #14
    mtbr member
    Reputation: yetibetty's Avatar
    Join Date
    Dec 2007
    Posts
    1,781
    I can't remember why I used Arctic Alumina, it was 3 years ago but I think it was the less electrcally coductive at the time. I havn't used it since exept for joints in housings.

    I now only use Artic silver 5 paste, great as you can align optics better and upgrade LED's.

    EDIT: Changed the siver 3 to silver 5, had to go get it out of the fridge and read the lable.
    Last edited by yetibetty; 07-13-2009 at 08:51 PM.

  15. #15
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    cdcdcd
    You are throwing up some interesting stuff with your poking around
    and thanks for taking the time to do it .

    Folks me included used Arctic Alumina as it was cheaper and also not recomended for fixing the Maxflex driver to its heatsink , as this was a diy light and a way to get a killer light way cheaper than the off the peg stuff.
    This has changed quite a lot in the last couple of years as the quality of the lights people are making has increased to rival the top stuff if not surpass .

    Do you read and post on Candlepower forums
    if not then can I suggest you post your findings and questions over there also as they may be of interest and there is a huge knowledge base for the workings of leds and thermal managment .

Members who have read this thread: 0

Posting Permissions

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

THE SITE

ABOUT MTBR

VISIT US AT

© Copyright 2020 VerticalScope Inc. All rights reserved.