FLATPACK - Dual MC-E design- Mtbr.com
Results 1 to 20 of 20
  1. #1

    Reputation:
    Join Date
    Jun 2009
    Posts
    82

    FLATPACK - Dual MC-E design

    I was busy last weekend, and hacked the following bike light from a rectangular block of aluminium I found on the workshop floor.

    Compact external dimensions are 57x35x42 mm, and the bare housing weighs 80 gms, though that will be reduced. I affectionately call him "flatpack". Within the front cavity reside two Cree M-bin MC-E emitters, mated to Ledil square optics, presently one type CMC-RS and a CMC-SS.




    More pictures follow, showing the constructional sequence, and eventually there will be the results of electrical measurements, thermal measurments, beamshots and who knows what, and all the while I'll be raving about why I designed this thing the way I did. For example, note the simple yet functional mounting points on each side, designed for thermal transfer to the mounting brackets (as yet unmade), and providing a fast and easy method of pivoting the lamp to adjust beam angle. OK, let's look inside the front cavity. The picture shows one of the leds being epoxied to the housing using Arctic Silver thermal epoxy. Recesses have been milled (1mm deep) below where the led pins sit, and 1.2mm diameter holes have been drilled through to the rear cavity to take the connecting wires. Note the jig (detailed photo later) used to accurately locate the led, and the two jig screws to press the led firmly onto the housing while the epoxy cures.




    Mounting the leds in this way was a piece of cake, and I could mount 100 leds using this method and guarantee that every one of them was perfect. I estimate the bond thickness to be essentially equal to the surface roughness, less than 0.02mm, guaranteeing an interface with negligible thermal resistance. The M2.5 tapped holes for the jig screws serve another purpose as well - more about that later. Here is a shot of the jig block, made from perspex (acrylic) so I could observe the epoxy being forced out from under the led, as well as visually make sure that excess epoxy did not foul the electrical pins.





    Here are both leds fully mounted and bonded. Note the precise location with respect to the holes for the connecting wires. Each MCE is wired in series parallel, so there are 4 connections per led, and 2 pins are electrically linked together at each connection.





    Here are the leds after the connecting wires have been soldered on. This soldering of the connecting wires was a breeze. Taking the time to make the jig and do everything properly was well worthwhile. Each led is 2S2P, and the two leds are wired in series, so overall led voltage is (about) 3.2x4 = 12.8V.




    Before moving to the rear end, how to seal the front end? Square optics are tricky in that way. When I decide exactly which optics to use, both optics will be glued together using silcone sealant- that will seal the small gap between the two optics. The picture below shows how the inside front of the housing has been machined 1mm oversize, forming a 1x1 mm groove around the front face of the optics. I then have 2 options. A thin-section O-ring can be pressed into this groove, forming a highly water resistant seal. For a fully watertight seal, the groove can be filled with a fine bead of silicone sealant. The optics scratch easily, explaining why they are recessed a small distance inside the housing.




    I have borrowed ideas from many of the lights here, and in some ways this is a miniature (2-MCE) version of Troutie's "insane" 4-MCE light. However, I thought that having the controller sort of hanging off one end was a bit ugly, and complicated the mounting brackets as well, so have instead placed the controller cavity at the rear as below, with many advantages for me. To be fair, Troutie was constrained by needing to fit in a large circular Hipflex controller.



    Rear cavity dimensions are 44x23 x 20mm deep. Most of the Taskled drivers would fit, but I could not live with any of them on account of the primitive user interface. The (relatively) large volume and rectangular format are ideal for accommodating my preferred option of a homebuilt controller. Also, the large rectangular rear panel is ideal for locating switches and/or pots which give greater power and flexibility for the user interface. In particular, I will use a small pot for intensity control, combined with an main on-off toggle switch, and another toggle to instantly flick from the intensity set on the pot, to a "low beam" setting of maybe 300 lumens. That's essential if you want to actually use the 1600 or so Lumens in the city or on the cycle paths. The rear panel is not made yet, but that's a minor job. It will be screwed to the rear of the housing (tapped holes shown), and weatherproofing will be with a thin gasket of paper or silicone adhesive.

    The four M2.5 holes visible which were primarily for use with the led-mounting jig double as a means of pressing out the optics. There is also a 1mm diameter hole just visible, drilled into the housing right under the mounting surface of the led. During testing, a small thermocouple inserted in this hole will be used to measure the temperature difference between this point, and the outside surface of the housing.

    THERMAL DATA

    The light was powered from a Lab style constant current power supply, in still air. The temperatures plotted below were measured on the outside of the housing. A second thermocouple was inserted into the small hole drilled into the housing, under the mounting surface of the led, which is visible in the pic of the rear cavity. At full current of 700mA/die, the temperauture measured here was only 3.0 DegC higher than on the outside of the housing.





    At 350mA/die, the light could be operated indefinitely in still air, with the case temperature rising approximately 50 DegC above ambient to 72 DegC. The rated Lumen output at 350mA/die (M-bin) is 460x2 = 920 Lumens, which is probably more than anyone would need when running for long periods in still air. At 45 DegC ambient, the case temperature would be 45+50 = 95 DegC, which would still not damage the leds. I conclude that the light could safely be left at 350mA/die for any length of time, at almost any ambient temperature. That said, the led efficiency drops markedly at these high temperatures, and the driver would be severely thermally stressed as well, so the cooling is not really good enough for regular and prolonged use at 350mA/die. That probably does not matter when used aa (moving) bike light.

    At full rated current of 700mA/die, the temperature rose to 95 DegC in 13 minutes, at which point I decided to wind the current back to 500mA/die, after which the case temperature stabilised at 90 DegC. At this current level, the case temperature would eventually rise to around 100 DegC above ambient, which would probably destroy the leds. To be precise, at full current, the led junction temperature will be around 30 DegC above case temperature. So, for Tambient=25DegC, junction temperature would eventually get to 25+30+100 = 155 DegC, which is above absolute maximum rating of 150 degC. Personally I'm confident I would never let this happen, but to be bulletproof, overtemp protection would be required.

    I found it interesting that the temperature differentials across the aluminium casing were almost negligible, as in only a 3 Degree differential between the hottest part of the case where the led is affixed, and the coolest part on the external surface. Arguably that means that the housing is overdesigned in terms of the mass and thickness of aluminium. A substantial mass of aluminium acts as a buffer to limit the rate of temperature rise in still air, of course, but to me that is a bit like arguing that we should add lead masses to the rear of the bike to improve traction uphill! Yes, it works, but there must be a smarter way. As it stands, the bare housing weighs only 80 gms and the thermal performance seems reasonable for such a lightweight and compact design, so maybe I should be satisfied. Nonetheless, I can't help but realize that the 80 gms of aluminium is not optimally used, as the temperature differential across the housing is disproportionately small compared to the temperature difference between the housing and the air. For a given mass of 80 gms, the thermal performance would be better, for example, if the fins were made 3 times thinner, and 3 times longer. I kinda knew that when I designed and built the thing, but machining the longer, thinner fins would be more work, and I would have needed a larger block of aluminium. Even so, I am tempted to make the existing housing and fins thinner for an ultra lightweight 1600 Lumen light, as these measurements tell me that this will have almost no effect on the steady-state thermal performance. With overtemperature protection the lowered thermal mass won't be an issue, as I'll never need more than 900 Lumens coninuous rating in still air. 50 gm bare housing mass would be nice, and is certainly possible.

    I also have plots of forward voltage versus temperature, but that's enough for now.

    It would be interesting to repeat the above plot with the same housing, except painted (or anodised) black.

    Have also built and tested a very simple linear regulator which works well, but my lust for efficiency demands that eventually I'll home-brew a switchmode design.

    Also need to machine the mounting brackets so I can do some riding and see how it all works in practice!!

    Colin
    Last edited by cdcdcd; 07-21-2009 at 02:17 PM.

  2. #2
    aka RossC
    Reputation: ocean breathes salty's Avatar
    Join Date
    Oct 2006
    Posts
    936
    I love it. The jig is a real artwork!

  3. #3
    mtbr member
    Reputation: tamen00's Avatar
    Join Date
    Mar 2004
    Posts
    1,025
    very cool ideas, I like the jig also. Now, just waiting for all the tests on it... thermal, electrical, all the stuff that has been talked about in the other threads. I am especially interested in the thermal performance and the measurements... What driver did you end up going with and what efficiency are you seeing?


    Nice light!

  4. #4

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by tamen00
    very cool ideas, I like the jig also. Now, just waiting for all the tests on it... thermal, electrical, all the stuff that has been talked about in the other threads. I am especially interested in the thermal performance and the measurements... What driver did you end up going with and what efficiency are you seeing?


    Nice light!
    I suspect I was editing a bit more info re drivers onto the end of my post while you were writing. There is no driver as yet - I started this project less than a week ago! I'll use a constant-current Laboratory power supply for the thermal and electrical measurements of the light, but my next job is to knock up a very quick and easy linear regulator just so I can get up and running riding the bike.Then, it will take a bit longer to develop a high-efficiency switchmode regulator, but my previous experience with switchers suggests I should get >95%. This is not a "whizz-bang" light, but I hope it is a"nice" light. Thanks.

  5. #5
    Spanish biker
    Reputation: msxtr's Avatar
    Join Date
    Dec 2006
    Posts
    732
    Super nice light!!!!!!, why you don't have used the MC-E leds with stars??

    Greetings - Saludos

    msxtr
    Warning!!! my english is very very bad, sorry.

    Easy DIY led light1
    Easy DIY led light2

    The Beast!!!

  6. #6
    mtbr member
    Reputation:
    Join Date
    Jan 2008
    Posts
    8
    Impressive design!

    If I may make a suggestion. From my overclocking days I learned that it's a good idea to flatten and polish the thermal mating surface(s) in order to keep the thermal interface material (in this case AS epoxy) as thin as possible, and to provide as much direct contact as possible.

  7. #7
    mtbr member
    Reputation: yetibetty's Avatar
    Join Date
    Dec 2007
    Posts
    1,781
    Very well thought out, good luck and I too look forward to the the thermal test results.

    Hope you get better results than I did

  8. #8
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    Looks nice CD, can't wait for test results. I liked the fixture for glueing the LED's. I made my fixture out polystyrene so the thermal epoxy wouldn't stick to it. I've got 3 MCE's and the heat measurments were the show-stopper for me. I was pleased to see I was drawing 24.36W and putting 22.25W into the LED's. Only about 8% used in the regulator. With 88WH in battery, I won't lose too much. The good thing is, I'll be able to warm my hands on the rides this winter. lol James

  9. #9

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by msxtr
    Super nice light!!!!!!, why you don't have used the MC-E leds with stars??

    Greetings - Saludos

    msxtr
    Yes, I could have used stars, but thermal performance is slightly better by gluing the led directly to the casing. I like to do things the best way, even if the performance gain is slight. Ironically, one of the MCE leds was on a star, and I had to unsolder it.

  10. #10
    mtbr member
    Reputation: troutie-mtb's Avatar
    Join Date
    Sep 2007
    Posts
    3,062
    That is looking pretty small and neat
    Looks plenty well cooled for riding along and I am looking forward to your results when you get it running .

  11. #11

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Quote Originally Posted by mlitscher
    Impressive design!

    If I may make a suggestion. From my overclocking days I learned that it's a good idea to flatten and polish the thermal mating surface(s) in order to keep the thermal interface material (in this case AS epoxy) as thin as possible, and to provide as much direct contact as possible.
    That is true, although my reading about structural epoxy bonding, for example of airframe components, suggests that it is important from a mechanical viewpoint to maintain a very thin epoxy layer, even if it is only as deep as the imperfections in the surface finish. To take this argument to the absurd extreme, if you literally squeeze out every last nanogram of epoxy then the components will no longer be bonded.I also considered the other thermal resistances which are always going to be much larger, and you reach a point of diminishing returns. Calculation shows that a 0.1mm AS epoxy gap will result in a temperature difference between led and heatsink of around 8 DegC, which is excessive IMHO. However, once you get the gap down to 0.02mm the temperature differential is less than 2 DegC, which is negligible compared to the 28 DegC differential from semiconductor junction to solderpoint, not to mention the differential between heatsink and the surrounding air. But I agree with you. It is important to keep the epoxy layer thin. Sounds like we are both perfectionists.

  12. #12
    mtbr member
    Reputation:
    Join Date
    Jan 2008
    Posts
    8
    I do have the bad habit of taking the absurd to the extreme!

    And i wasn't trying to suggest that your work was imperfect. Quite the contrary, I think the design is impressive. I was just thinking (typing?) out loud that if you were going to the trouble of removing the LEDs from their stars for the purpose of better thermal transfer, you might want to put some time into making the mating surface smoother than raw tool marks.

  13. #13
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    +1 on the polish job. I have a Intel Quad-core that dissipates 120W. I polished and flatened the micro processor and the HS to a mirror finish. Then AA'ed the HS into place. This dropped the processor temp by 13 degrees. I also flatened and polished the stars and HS for my light. A piece of glass, some wet/dry sandpaper, some wet sanding is all it takes. Prolly spent 5 minutes.

  14. #14
    mtbr member
    Reputation: yetibetty's Avatar
    Join Date
    Dec 2007
    Posts
    1,781
    Quote Originally Posted by OldMTBfreak
    +1 on the polish job. I have a Intel Quad-core that dissipates 120W. I polished and flatened the micro processor and the HS to a mirror finish. Then AA'ed the HS into place. This dropped the processor temp by 13 degrees. I also flatened and polished the stars and HS for my light. A piece of glass, some wet/dry sandpaper, some wet sanding is all it takes. Prolly spent 5 minutes.
    Sorry cd but me too
    Attached Images Attached Images

  15. #15

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    +1 on the polish job. I have a Intel Quad-core that dissipates 120W. I polished and flatened the micro processor and the HS to a mirror finish. Then AA'ed the HS into place. This dropped the processor temp by 13 degrees. I also flatened and polished the stars and HS for my light. A piece of glass, some wet/dry sandpaper, some wet sanding is all it takes.
    Sorry cd but me too
    The thinner the epoxy bond, the better, thermally speaking. Agreed.

  16. #16
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    Cd, as you mentioned before, the thinner the epoxy bond the better. I worked on F-16 fighters for many years. They are glued together using an epoxy made by Hysol. The metal is preped & cleaned and the parts are clamped until the glue sets. The metal (aluminum in F-16's) will give before the bond. So to recap: smooth, clean, flat all are important. If we are going to bond pieces together, we don't want the surface too smooth (like 180/220 grit sandpaper is as fine as we want). Use laquer thinner and clean clothes to get ALL the oils and greases off, Wipe all parts off several times. Wear latex gloves to prevent contamination from skin oils. DON'T get laquer thinner on the LED dome! Clamp parts with lots of pressure. easypeasy, James

  17. #17
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    Well being that I'm old and forgetful, I need to add one more point. The desired epoxy bond is about .002". Later, James

  18. #18
    mtbr member
    Reputation: piesoup's Avatar
    Join Date
    Feb 2009
    Posts
    455
    Quote Originally Posted by OldMTBfreak
    I worked on F-16 fighters for many years.
    I worked on the RAF Merlin helicopters, they are bonded together too, admittedly they hae a composite construction though. I still find it amazing that they are just 'glued' together! Never knew the F16 was too!

  19. #19

    Reputation:
    Join Date
    Jun 2009
    Posts
    82
    Well being that I'm old and forgetful, I need to add one more point. The desired epoxy bond is about .002". Later, James

    Yes, that is the point I was trying to make before. From a thermal point of view, the thinner the bond the better, right down to zero thickness and metal-to-metal contact. However, from a structural point of view, it is desirable to have a thin but non-zero bond thickness, and that is why I would personally hesitate to polish the heatsink to a mirror finish. I would be less worried about the structural integriry of the bond if the led and heatsink were made of the same material, such as both aluminium or both copper, but in fact the led solder point is copper and the heatsink almost always aluminium, and the differential thermal expansion will mechanically stress the bond. Should the bond fail mechanically the led is toast. My approach therefore is to intentionally have a non-mirror surface finish, such that the effective bond thickness is set by the depth of the surface imperfections. Thus, I can (and did) clamp the led to the heatsink with substantial force while the epoxy cures, confident that the structural integrity will still be good due to small but non-zero bond thickness ensured by the surface imperfections. But this is just my personal preference.

    In practice, I suspect it makes little difference what you do, just so long as the bond is thin, say less than 0.05mm.

  20. #20
    mtbr member
    Reputation:
    Join Date
    Apr 2006
    Posts
    596
    When I am polishing the HS or the stars, for our lights, I don't use too fine a grade of sandpaper. On the computer, I polished both the microprocessor & HS because the HS isn't bonded. Just heatsink compound applied between the pieces. Yep, .05mm is about .002".

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.