Also, once we have gerbers I would have an Oshpark set made to test but I think it will work best regarding heat with a board as thin as possible that has thicker traces.
Pcbway is one of the fabs I checked out this morning.
Not much price difference between 100 and 250 pcs $127 for 250 shipped DHL
24 mm x 24 mm 2 layer .4mm fr4
Enig
White mask w/black print
1oz(couldn’t get auto quote to insert 2 or 3 oz.)
Other places could do metal filled micro vias which would be better but need a design to get a quote.
There are also fabs that have plants world wide.
As copper thickness increases, so do trace width and spacing minimums. Anyone have an idea on best compromise? It seems to me that in this case we could be more generous with spacing to maximize the copper on the board. 2-3oz?
My eagle might have flew overhead and dropped a rough board in my lap. Except oshpark min board edge slot is 2.54mm. The pegs, of course are 2mm. Oshpark’s fab can do smaller internal slots but they are not guaranteed and may not be manufactured. For a test board at least, you could do only 2 holes for the pegs. 2 for Khatod left and right, 2 for Carclo top and bottom. Then snip off the tips of the other 2 pegs.
Look for min supported slot size for the fabs. Though they may not list it in their specs. 
Copper thickness, power traces on some boards use exposed copper without soldermask that gets covered in solder to bulk up it’s current handling. We could use that for narrow sections of the traces.
Question for the fabs, will a board with thicker copper also receive thicker plating of the vias? Guaranteed minimum?
It seems you can only select 2oz for boards 1mm or thicker.
I looked at pcbway because they had an estimator but I think well have to look around a bit more. Rushpcb seemed pretty advanced with silver filled laser drilled micro vias and lots of board choices. I’m slightly out if my depth in knowing just which fab to use since I don’t know most of the material choices to choose from.
Not nearly as cheap but I can make pretty much any custom DTP copper mcpb by machining traces into copperclad board with any thickness of copper (4oz, 6oz, 10oz, etc.) and then I just machine pockets through the board and machine bosses onto a copper heatsink. I then either reflow or adhese the machined trace board to the copper heatsink and voila, instant custom DTP board. I could then epoxy paint the boards for insulation and re-machine through the paint for the pads. Thick enough copper traces will allow this because I would only have to machine through the paint .001” or .002” for example.
It would not be $1 a board though it would be more like $12 a board.
This one can be used with either optic and arranged in 4s, 2s2p, or 4p:
With a description like yours above Hoop we need pictures of this happening or at least of something being made to ensure all of us here that you are indeed not from another planet.
Any chance I could drop in for a coffee sometime. Please.
:bigsmile: You’re funny MRsDNF. I will set aside the time to make the physical version of that model within the next few days. I only have 1oz copperclad on hand but it will suffice to prove the viability of the process.
And yes, I’m all about the coffee.
I took measurements of the lights I have.
Solarforce L2M - 22.5mm ID
UltraFire WF-502B - 24mm ID
Convoy M1(1st gen) - 24mm ID
These are all hosts in which this pcb or the host could be adapted to fit. Please post some other examples if you have them.
That would be an awesome board hoop and almost certainly a more capable one at high currents. It’s high for getting a bunch but I’d be interested in one anyway just for the uniqueness of it. Maybe start a sign up thread with me at #2. 
The fr4 board I have in mind would be a budget minded way to get something that I hope would be supplanted by a real copper board in the future.
Ultrafire WF504B 22.6mm.
Solarforce L2M 22.53mm
Yea Scott, the price of your boards can’t be touched. Now that’s designing for a budget.
I don’t have anything to add to the project. But I know I would be a customer for a bunch of them. I have about 100 or so 502B’s not that I want them all to end up quads but I would take at least 25 of them.
I certainly don’t need that many myself but it seems doable for both Hoops’ version and a budget model.
I've already got an Oshpark board that I threw together last week, but as usual it needs to be tested before it's ready for "showtime".
Great! I have a feeling from what Pilotptk did with Scaru’s board and the 35mm driver on board that it needs to be thinner than Oshpark can make them but if you have a fab for .4-.6 mm count me in. I also have a few extra Khatod optics if you’d like to see for yourself.
Nice project. I can’t wait to see what you all produce. $12 a board for thick copper in a quad, that might not be so bad. Especially with all that heat as I’d have to use XPLs.
The Oshpark boards are so pretty, I really like them a lot. But they aren’t direct thermal path.
Hoop’s design looks nice. I wonder if that could fit into a convoy M1 with a copper spacer and a tight Katod optic array.
It certainly could. The M1 and 502B both are 24mm ID down inside but the M1 flares out to a larger diameter at the bezel to accommodate a bigger reflector. It would be a good candidate for the 35mm PL 251106.
The Oshpark board and any production run of thinner fr4 boards would be DTP just not solid copper thermal pads but instead solder filled vias. At this point silver filled is also a possibility but my uneducated guess is that it would not be a budget option.
I think that if done properly the solder filled bias work pretty well, but are more work than most would like. You really need to fill them fully, which always spills out onto the backside, which then needs to be made flat again. The ideal thing would be to have a 4 oz copper board made because you really lose out on the heat spreading capability on the backside, but at that price and volume point you might as well go all the way to an MCPCB. I don't expect that these will perform as well as a DTP copper MCPCB in a direct drive scenario, but at 2A-3A per LED I think they might be a viable solution.