Soldering a MCPCB vs Thermal Adhesive (test results)

Flashlight Modding and DIY Parts
1

I’ve always wondered if soldering a Noctigon MCPCB to a pill would be worth the trouble, so I tested it. I made 4 identical Convoy M1 builds…2 attaching the MCPCB with Fujik and 2 attaching the MCPCB with solder. The builds use a 20mm Noctigon with a U2 1D (the chart shows U2-1A, but I actually used a U2-1D) driven at 3.8A installed in a Convoy M1. The pills were mounted (screwed) into the host with thermal silicone on the threads. I ran each light, collected the data and averaged the 2 with Fujik and averaged the 2 with solder. I then plotted the results…so here it is:

!!

2

Bout 50 lumens difference? I wouldn’t bother soldering :slight_smile:

3

Very nice tests. I wonder if there’s a huge difference between fujik and the Arctic or other higher cost adhesives.

4

While there isn’t much difference in lumens the soldering helps the plateau after 10 min.

5

Nice test. Thank you. To me it shows that soldering is the way to go since the unsoldered one is still dropping quite fast at the ten minute mark and the other one has a less steep angle. Maybe the host had saturated?

Can I ask if the surfaces were lapped / polished?

6

Great test. Thanks for plotting it for us.

7

The Noctigons were sanded with 600 grit (to remove the die cut burrs), then with 2000 grit wet. The pill platform was polished with 0000 steel wool using a drill press. There were still some small ridges left in the pill, but these are very difficult to remove.

When I use silicone to apply a MCPCB, I smooth the surfaces, apply silicone, mate the surfaces with pressure, disassemble to check for coverage, reassemble and clamp. I then heat to 125ºC to thin the silicone. I increase clamp pressure at about 5 and 10 minutes. This squeezes out and thins the silicone layer ensuring good metal to metal contact.

I think soldering is easier on some lights…I have trouble with centering when I try to solder a P60.

8

Knowing you (or at least your uncompromising style when dealing with thermal management) I suspected they would be polished :-)

Except for heating them (thank you for the tip about heat thinning the silicone) our modus operandi seems similar.

And about soldering P60. I found that a thin band of aluminum from a soda can around the inner edge helps me tremendously in centering the mcpcb and keeping it centered. It was always keeping them centered that bugged me.

9

Thanks for the test and results. I’m glad you are motivated to do this.

Looks like you’ve demonstrated the absolute best case scenario for Fujik :wink:

10

Thanks for testing. Quite a difference.

I wonder how good thermal paste would compare. Fujick isn't really that great.

11

I read it as 70 lumens after 10 min. and still increasing. Maybe not worth much trouble at this current, but it could take more current safely.

12

Thanks for the test.

I’d say the gain is worth it if you’re maxing your mod - the graph spread will be much higher when doing 6A or so.

I do wonder too where the line plot of regular CPU grease (e.g Arctic Silver) with tightening screw would be… :wink:

13

Thanx for the tests! Yeah, I also wonder what AS5 and the like will do compared to the solder and Fujik :)

14

I would guess that electrically conductive glue with very high silver content might do as well as solder. Silver itself is a good conductor and corrosion resistant when protected from sulfur. It is a matter of how much contact there is between the particles.
It seems that how smooth and flat the star and pill are should also be important.

15

Hey pflexpro, it just hit me. You didn’t mention how much of a difference there was between your samples before you did the averaging.

Will you please tell us about that?

16

Interesting comparison, thanks. My personal expectation would have been that there was no difference at all if the fujik was applied the way you did in this test, but this proves (again, I noticed it before), that every obstacle in the heat path from emitter to air does have its (minor) influence, even if it is not the main bottleneck.

17

I’ve replaced the chart with all four runs and their average.

18

Again, nice work. Thanks.

19

Hey pflexpro, I'm assuming you still have to the lights and hardware you used to do this test. I see the test was done at 3.8A. Would it be possible to redo the test ±2.8A and ±3.8A to see what the difference would be between direct soldering and adhesive at different amp levels?

Thnx

20

I can re-run the tests but the driver will still be 3.8A -I can limit the input current to the driver to 2.8A -I think this should give a similar results as if I used a 2.8A driver. Let me know if you think this would show what you’re looking for.