If you make the shelf thick, and the holes for the wires just barely bigger’n the pads on the LED, you could maximise heat-flow through the critter. As drawn, the LED’s thermal path is just resting on that toothpick of metal which’d constrict heat-flow.
Once it’s mounted, quick tack-soldering of wires to the LED could be done before the center-pad’s solder melts.
Yep, that was my main concern, that the restriction of the two ovals would negate too much of the benefit of getting rid of the MCPCB. The smaller the holes and thicker the shelf, the more difficult the soldering job becomes.
I don’t think I’m steady enough to pull that off, but it’s a great thought.
Treat the pill like a hotplate. Heat it with, say, a hot-air gun, and once the LED’s soldered to it, cool it with a fan.
This seems pointless just reflow a good mcpcb onto your piece. No paste needed that way either… people have been doing this forever
Okay, more ideas struck me. How about something like this:
No big oval holes in the copper right next to the LED. There's a small tombstone boss in the center of the copper for the LED to solder to. The green part is a layer of Kapton tape. The two orange pads are the cathode and anode solder pads. Here's a section view:
Sorry that Fusion screws up all the colors in the section view. I hope you can still make sense of it. See how the copper piece is still monolithic, and the kapton is just insulating the two pads? I'd need some sort of high temp adhesive to attach the pads to the Kapton. With some careful temperature control, the whole thing could be reflowed without coming apart.
Again, the copper piece is going to be way bigger than what's shown. It'll be a big monolithic heat sink that the rest of the flashlight screws onto. We're just looking at the interface of the LED and the copper because that's where my questions lie.
Edit: I just discovered that there's a double sticky Kapton tape! It seems like that would work to hold the two pads down during the reflow.
From what I understand, solder isn’t the best conductor. I read on some old BLF post that soldering the MCPCB to the pill is worse than lapping the surfaces and screwing the MCPCB down to the pill. Surely eliminating the MCPCB and going with a monolithic center section would be even better, right?
That’s true, but the only way for a TIM to be better is to use liquid metal.
A soldered down MCPCB would still be a better idea.
However, the idea has already been executed, but only with watercooling.
Idk I’ve never felt like that was a bottleneck in any of my builds. Especially what I can only assume is a single 3V LED you are doing this for. I don’t want to discourage you but unless you have unlimited budget for weight you should focus on outside host and driver/LED efficiency to make bigger gains than whatever thermal paste transfer bottleneck you are trying to solve.
Seen it done a few times. Back before DTP MCPCB became available.
A basic insulated MCPCB with a drilled out central pad and a copper or brass pill with a nub machined to fit it.
I’m going for increased turbo times from a FET driver on a small EDC light. And unless I misunderstand things, removing heat is exactly the main bottleneck when trying to increase turbo times. Also, if I decide to build something really extreme someday, this type of attention to detail is exactly what that’ll need.
The LED shelfs on even our budget lights are good now but once it gets saturated you need surface area (fins) to make any more headway without the light just being a brick of copper.
That part is way harder to do right. We already get the heat off the LED good enough. Work on a good design to get it out of the light
I have made a copper heatsink similar to your design many years ago. It wasn’t my idea, I just followed what someone else had already done. I saw it done a couple of times over at the other flashlight forum long ago.
It does work well. The only fault was that the wires being directly soldered on to the led was a little hard to do and was fragile while trying to assemble everything.
This was before any of us used flexible silicone wire which would take some strain off the connection. I tried to find somewhere that someone had posted up how it was done. I know I have seen it a few times years ago but I can’t seem to find the exact method posted anywhere now. Once the copper direct thermal path Sinkpad was invented there was no need to do it this way any more. It might be better than a Sinkpad if soldered directly to a copper heatsink but the copper mcpcb’s are just so much easier to work with and easier to change out if you want to change something. More modder friendly I guess.
I used a propane torch to heat the heatsink to reflow the led directly to it while held in a vise. It was a crude uncontrolled reflowing method that would cook a emitter every now and then. Back then, gaining 100 lumens from better heatsinking was well worth the effort.
.
I would have to agree with contactcr, thats where the bottleneck is.
Seems to me that building this into a big monolithic center section covered in fins is as good as it can get. Or maybe a press-fit of something like this into a larger aluminum heat sink. But other than that, what else can be done?
How about this. Make a heavily finned light then create a ball bearing sleeve around the handle so when you want turbo you give it a spin and go to town. You can add a latch to keep it in place for lower modes.
Practical and fun and still mostly water resistant!
Why?
OK, that’s different. Instead of having a fan, spin the heat dissipation fins for more air circulation 
I think you’re on the right track for high wattage applications. Interestingly, Lambda Lights used to mod and resell high amp LED maglites with the same mod. He had to heat the entire chunk of copper (around +2 pounds as I recall) in order to flow the emitter onto the gargantuan copper pill. There were technical challenges he had to overcome before they would adhere. Here are some of his creations. I owned several before reselling.
You can try another way.
Regular pcb (fr4 or alu based) with central pad hole, and copper plate with extruding pin of same shape. This way have no user benefits in comparation with dtp pcb, but making fr4 or alu core pcb of any shape and design is widely available (while dtp pcb is not).
Brilliant! I’ll make my pads out of standard double sided PCB. I can machine pockets into the copper base material for the pads, and even reflow the pads onto the copper. It’s basically making my own MCPCB, but the MCPCB is a huge piece of copper, with heat sinks machined into it. I’ll make sure to cut the pad reliefs a few thousandths deeper than needed, so the center section (the tombstone that the LED mounts to) is proud just a tiny bit. This will ensure the thinnest film of solder at this critical point.
I’ll model it up and post it in the morning.
Holy F! I remember those lights from back in the day! Like I think I even bookmarked that exact thread in my early days of lurking. LED technology sure has progressed in a decade.