1

I am making my own LED bulbs (for wedge-base sockets in a low-voltage application). A key construction detail requires me to cement a small (roughly 3/8" x 1") piece of FR4 circuit board (which plugs into a wedge-base socket and holds a couple of small components) to an aluminum "star" (a thermal substrate for the LED emitter itself):

https://www.mouser.com/ProductDetail/Bergquist-Company/804087?qs=jQRjkUoUCJebprw0Kn9Vjw%3D%3D

The FR4 will be perpendicular to the star, with the narrow 3/8" edge of the FR4 against the back of the star.

enter image description here

The Osram Oslon SSL emitter will be running at up to a watt (but usually closer to 3/4). They will operate continuously for up to 10 hours at a time. The maximum spec'd junction temp is 135 C.

My question, what is the best method for making this attachment? I have tried JB Weld epoxy (two part), 100% silicon caulk (specifically GE clear Silicon II), and intend to experiment with Sugru. The epoxy is quite strong, but I have managed to break the connections when pulling bulbs out of sockets when they have gotten stuck. The silicon has the advantage of flexibility, but I don't think it's as strong; it's also easier to assemble, as it has enough stiffness to hold the FR4 in position while it cures, whereas the epoxy requires bracing. I've never worked with Sugru.

Are there other methods I might consider? Factors I haven't considered?

winny
  • 13,064
  • 6
  • 46
  • 63
RustyShackleford
  • 351
  • 3
  • 14
  • 1
    Do you have any specs for temp rise of junction e.g 50’C and done any calculations on Rth . Most Plastics including silicone, epoxy , iron particles in JB weld , FR4 make great thermal insulators, but thinner is better. also PU subfloor structural adhesive on edges is good. Generally 99.9% silver epoxy is best with thinnest layer and structural support to prevent warp. – Tony Stewart EE75 May 26 '18 at 19:38
  • 1
    I've recently used [JB Weld: Plastic bonder](https://www.amazon.com/J-B-Weld-50133-Structural-Adhesive/dp/B009EU5ZNO/ref=sr_1_4/146-3955006-1385744?ie=UTF8&qid=1527366531&sr=8-4&keywords=jb+weld+for+plastic) to mount nylon standoffs on a zinc plated steel base. Very mild flex and a rock solid bond. Would imagine it would work well for your needs, but can't say for sure. – Phil C May 26 '18 at 20:35
  • @Tony, I am going to be running less than a watt, and preliminary results (very adhoc, admittedly) show the star not getting too hot. But it's a good point that I need to avoid insulating the rear of the star too well. Those thermally conductive epoxies look a lot better, but very expensive, so I'll only use if I need to. – RustyShackleford May 26 '18 at 21:26
  • 1
    Part of the problem is the thermal expansion and contraction of aluminum with temperature. I was able to bond a aluminum bracket to a stainless steel box only with RTV 262, which has no ammonia in it. All types of epoxies and cements failed the heat/freeze test. I do not have a better opinion than RTV. –  May 26 '18 at 21:27
  • @Tony you're not suggesting PU subfloor adhesive is particularly good for thermal conductivity, but good strength, right ? What do you mean "on edges" ? – RustyShackleford May 26 '18 at 21:27
  • 1
    My expectation is that epoxy will fail at high temperature. However, it may work a bit better if it is specifically designed for high temperature and if you are able to apply a controlled post cure heat treatment. Silicon seems a good choice. You MUST use a silicon designed for PCB's. There is also double-sided acrylic adhesive heatsink tape. Whatever you use, you may need to perform some surface treatment first. Any film residue on the aluminum or FR4 can cause the bond to fail. – user57037 May 27 '18 at 06:33
  • I do also think that edge gluing FR4 is not a good plan because there is not much surface area to work with. So you may want to rethink the whole approach. – user57037 May 27 '18 at 06:34
  • 1
    @RustyShackleford correct. I was not suggesting PU for thermal bond but structural strength for CTE shear and warp forces. The silver epoxy only needs to fill the pits of non-coplanarity clamped to remove all air gaps so very thin. CPU use silver and ceramic filled grease under extremely flat surfaces with strong structural spring force to do the same. But here structural weakness is evident. – Tony Stewart EE75 May 27 '18 at 14:43
  • @mkeith, I am not just edge glueing (should've been clearer about that in OP), rather, there would be a cone-shaped fillet of adhesive, that is maybe 3/8" diameter on the back of the star and then comes up about 1/4" on the sides of the PCB – RustyShackleford May 27 '18 at 20:37
  • @mkeith, by silicon designed for PCBs, you mean "neutral cure" as opposed to the usual "acid cure" ? As described in this comprehensive post: https://electronics.stackexchange.com/questions/18525/what-kind-of-glue-should-i-use-for-pcb-mounted-components-to-avoid-vibrations – RustyShackleford May 27 '18 at 20:48
  • 1
    Yes. neutral cure. But more than that, in the absence of expert knowledge, I would use something made by a brand name adhesive vendor where the product literature specifically mentions application to PCBA's. It is important that there are no conductive additives, for example, which might short out circuitry. But ultimately, because of the way you are doing this (on edge), I really think you should reconsider. Maybe make a small metal bracket to join the board to the star. If you rely on a fillet, the fillet material needs to have strength. This argues in favor of epoxy and against silicone. – user57037 May 27 '18 at 21:01
  • 1
    There are tradeoffs and you may have to test several different solutions. Also, you need to examine how things fail. Did the bond fail right at the board/epoxy boundary or aluminum/epoxy boundary? If so, surface prep may solve the problem. If the epoxy failed (cracked in the middle of fillet), then you need to go to a stronger material (like aluminum). – user57037 May 27 '18 at 21:02
  • My failures with epoxy have been at the epoxy/aluminum boundary. The PCB has a lot more irregularity for the silicon to grip, and I will ultimately design my own PCB, so I can put in a big via hole, so there'd actually be a "bridge" of adhesive holding that bond. I haven't tried roughing up the aluminum yet (just cleaning and giving a quick pass with emery cloth). Again with my own PCB design, a little conductivity in the adhesive will not be a problem. The silicon really seems pretty strong, and I am leaning towards it. – RustyShackleford May 29 '18 at 03:45
  • If I did a small metal bracket as you suggest, I could of course screw it to my PCB design. I'd still have to cement it to the star aluminum. I'd have the advantage of similar materials, and a larger bonding area though. It would also help with the (lack of) thermal conductivity issue. – RustyShackleford May 29 '18 at 03:46
  • 1
    Can you make a hole in the star for the FR4 to go through so that you could put a pin through the FR4 to secure it? – Andrew Morton Jun 17 '18 at 16:24
  • I don't think I can fit a hole that big (without disturbing the LED emitters and the traces to the pads for soldering leads), but a small hold might could be useful anyhow, to secure an L-shaped bracket, and forget adhesives altogether. – RustyShackleford Jun 17 '18 at 20:36
  • 5si years on. How did thse work out longer term? Which adhesive worked best? How many did you make? – Russell McMahon Apr 16 '23 at 11:17
  • @RussellMcMahon I ended up taking a piece of 3/8" aluminum angle and cutting 3/8" wide pieces off. Connected to PCB with a tiny sheet-metal screw and epoxy'd to the LED "star". A wee bit more work, but seems very reliable, plus the little piece of aluminum angle probably adds to the heat-sinking. I could use thermal epoxy, but doesn't seem necessary. I guess I've made about 50. – RustyShackleford Apr 16 '23 at 20:57
  • Of course the little aluminum bracket and the aluminum star have very similar thermal expansion, so that's good for the epoxy joint. And the other connection is a sheet-metal screw. – RustyShackleford Apr 17 '23 at 04:01

1 Answers1

4

This should be able to be "designed" with a reasonable degree of confidence either from theory alone or with a little experimental help.

The key parameters seem to be heatsink temperature, forces involved and adhesive performance with the materials used.

Independent of the adhesive considerations, the LED mounting temperature has a max upper limit set by the LED temperature ratings. As a first approximation, modern lighting class Cree LEDs are rated at 105 degrees C operating. That's hot by conventional electronic standards - you will not usually want to operate equqipment continually much above that unless there is an extremely good reason to, and lower is liable to be wise.

You do not say what wattage LEDs you intend to use, operating periods and duty cycles, or target temperatures or overall size. These combined will control whether it is possible to build your 'bulbs' without means of exterior heat removal.

A look at the technical data sheets of a few epoxy adhesives show a wide range of temperature specs. Some basic epoxies 60 C max operating) would not work. Others with a 120 C max "maybe" rating will "maybe" work. High temperature epoxies withstand temperatures that would rapidly kill LEDs.

Examples:

  1. Here is a semi-randomly chosen epoxy technical data sheet (TDS).
    BOSTIK EPOXY BOND 5 MINUTE TWO PART ADHESIVE

They say

When fully cured, the adhesive can withstand boiling water for short periods of time.Heat resistance rating of 60ºC temperature continuously.

ie - it would perhaps meet your needs for LEDs guaranteed to be on for short periods (indicator, stop ...) but would fail in applications where temperatures rose to 105+ degrees.

  1. EVO-STIK HARD & FAST TWO PART EPOXY RESIN ADHESIVE STICK.
    Temperature: -40 C to +120 C depending on conditions of use

MAYBE!

  1. Resinstech RT323 High Temperature Epoxy Resin System
    Operating Temperature: -50 C to 200 C

Suitable.

  1. MG Chemicals High Temperature Epoxy Encapsulating & Potting Compound 832HT
    Constant Service Temperature -30 to +225 °C
    Max Intermittent Temp. 250 °C

Suitable.

Looking at TDSs for various silicone rubbers would allow comparison, but there seem to be adequate epoxies available.

Russell McMahon
  • 147,325
  • 18
  • 210
  • 386
  • Thanks, Russell, for responding. I have edited the question (just below the figure), to add some of the info you mentioned. 135C seems awfully high, but it looks right (the "absolute" max is 160C): https://media.osram.info/media/resource/hires/osram-dam-2495583/GW%20CS8PM1.EM.pdf – RustyShackleford Jun 17 '18 at 20:52
  • I have mostly experimental evidence now, which is some test bulbs I've made with silicon RTV and with basic JB Weld epoxy. The epoxy will break if I pull too hard, but I doubt I prepared the Al adequately (I gather I should wipe with solvent, buff with emery cloth, and wipe again). The RTV works better, but will break eventually. A gentlemen at Loctite recommended this 2-part polyurethane, and I've got a tube on the way: https://www.ellsworth.com/products/adhesives/urethane/henkel-loctite-hysol-u-05fl-urethane-adhesive-off-white-50-ml-cartridge/ – RustyShackleford Jun 17 '18 at 20:57
  • These test bulbs have been running up to a year. The star doesn't not get too hot to touch (highly scientific). But I believe it's correct that temperature is the big issue, for three reasons: 1. Adhesive fillet insulates star too much, allowing emitter to overheat. 2. Temperature weakens the adhesive directly (glass transition temp of U-05FL is only 48C, could be a problem). 3. Temperature expansion/contraction of materials weakens joint. – RustyShackleford Jun 17 '18 at 21:04