problem with this methods is that you must have a little “technical mechanik ” feeling … :nerd_face:
Tom, you might try a variation of something I’ve done in the past. penumbra made me think of this and it should be simple.
I have used the double sided 3M emblem tape for cars, the stuff they put the Ford oval or Chevy bowtie on the side or tailgate with. It’s a thin foamy grey tape, sticky on both sides. A dowel or aluminum rod with this on the end, chucked into a drill or drill press, will enable you to turn that pill section like it’s on a lathe. You can then sand, rotary, file, scrape, however to get the turning pill flat on the surface. It’s most tricky to set the part gently on the adhesive and spin it slowly to test center, reposition til it’s turning true then press it firmly onto the tape. The stuff is quite strong, I’ve turned 7/8” copper bar stock this way to make a driver pocket in the end. This is also how I turn down a copper star. Turn it relatively slowly with my cordless drill and use a cut off wheel in the rotary tool to trim the star down. Takes a few minutes and works like a charm.
A piece of wooden dowel, like 1 5/16” clothes rod, can be shaped to fit inside the pill on a table saw, setting the saw blade for height so the round rod will be close enough to sand lightly then press fit in, once that end is done, the other end can be likewise shaped to fit into the chuck of the drill or table saw. I prefer the drill for it’s variable speed. Have no idea why I didn’t think of this before!
Thanks, Tom E. I've been thinking I may eventually attempt modding the Y3, but the E-switch part is intimidating. You make it look easy. 
I'm big on chucking stuff up to a hand drill too. I keep a variety of PVC couplings, pipe, and Phillip screwdrivers around just for that. But I have had bad luck trying to lap that way. Unlike a lathe where you have a fixed blade that moves across the surface that is turning at different speeds (Center material is moving past the blade slow, outer material much faster), the free hand sandpaper tends to dig in deeper on the outside of the pill landing. I always get a hump (and I don't mean the good kind) that way. So I prefer to manually lap my pills.
I'm sure I just don't know how to do it right. Hoping someone knows what I'm doing wrong.
Ohh - I ordered the single cell Y3's, so no extra tubes. Didn't try it yet - won't do me any good without ordering extras though.
Thanks for all the tips on working the pill top! Ended up though doing it same way as I always do, sand paper strip over the back end of a metal chisel - found it to be smooth and flat. I worked it extra with 150 GRIT more on the outside than the center, but don't think it did much - got a pic I'll post.
The OP has been expanded, adding tailcap assemble pics and details on modding.
Again I sit here admiring your modding skills Tom E. Orsm work on the driver and led.
Thanx! Few issues:
- I put a healthy nice dab of thermal grease and saw none squeezed out, so my rough test for flatness seems to have worked
- the stock big plastic piece seems too high - I get a black hole that goes away at bout 12" or so, but shouldn't really be there for a domed LED
- little disappointed in lumens and throw, but could be the centering piece, and also that I'm using a XM-L2 U2 1D which I've seen have less output than a 1A
- The light is pretty useless right now with the driver in a funky state - hope I didn't kill it
Wish I had pics or could recall what I did not on the fully modded Y3 I previously did a few months back... The reflector hole is an odd size I can't find an alignment piece to fit - something like 8.14 mm I believe. I'm thinking I cut the stock one back a lot, than filed it down - can't recall.
Tom, you can write an honorary chapter in the book Steve and I are Co-Authoring, the title is “Forget Me Not—-What did you say your name is again?”
Very nice teardown and mod as usual. 
I like the idea of extending the soldering pads on the star with the small copper sheets, that way we can eliminate the soldering joint interference with the reflector base. Think this has been done by someone else before too.
I've done it before (copper sheet metal tabs), but pretty sure I've seen it done/posted about. In a surface where you have no room around the MCPCB, this won't work, but if you have the space, works well. O-L did the recent video about using these extensions, but put them over the wire holes in the MCPCB, then solder the wire directly to the bottom of the tabs, not the top. That method will work well for a HD2010 that doesn't have the extra space.
Sorry missed this Q. I believe I bought two a while back - one I used, one spare, so will update this light and test the difference. First time I posted the improvements measured - did pretty well, bout 5% or so. I also may try improving the focus - think it could be made better.
I’ve also cut the copper out from under the pad on the outer edge, soldered the wire to the bottom of the pad. Wasn’t easy, but….
I used to be big into overclocking computers. I am onto better things now. But when I was, tried just about every thermal grease known on the planet. Try this stuff if you haven’t already. http://www.coollaboratory.com/en/products/liquid-pro/
It’s all I used on video cards, cpus in the end… It’s all the same stuff but this actually works in heavy overclocks and holds up over time if you can find a way to bolt it down. Adhesive, I only used to use artic silver artic alumina parts a and b.
Anyways hope that helps. I had extensive experience in computer modding and using the diamond pastes, and all the gimmicks in tim you could find… Made my own sinks. But really tried them all. That stuff is real. The rice grain method is still the best way to insure no bubbles under the sinks too. I saw a vid old lumens did on a light where he still spreads it That’s really old school and doesn’t work too well. I can link a video here with proof. That coolabs needs pre prep though where you rub it into the surfaces premount and it has trouble coming apart on aluminum after it cures. But it works so well, practically soldered, and by the time you want to take things apart it’s time to replace it anyway.
And unlike an overclocked CPU with ~200 watts to deal with in a tiny 1x1cm die, in a flashlight there's no measurable difference in output or runtime or efficiency or anything else that can be measured when comparing the best thermal paste against cheap white silicon-based paste.
That coollab stuff sounds interesting. What's the rice grain method?
My main problem lately though is getting a pill or shelf for the MCPCB that is actually flat. You really need that first I would assume before using something like this grease.
Presumptively, the rice grain method is to use a small bead of the thermal paste about the size of a grain of rice, in the center of the application. Then bolt the device down on top of this allowing pressure to squeeze the paste into a very thin layer. No bubbles. Am I right?
CC - I hear you, and recall that published test on the toothpaste vs. AS5 results (
), but, if it's true, this stuff isn't rated at 10% or 50% better than AS5, it's rated 9 to 150 times better (their spec), close to using solder. For me, I'd like to get solder level results for something that can be easily taken apart - this is all wish-list right now, because I still would have to start with a flat surface, and maybe it will never be achievable without re-building the pill of any of these flashlights anyway - seems like the $100 (or more?) lights aren't much better than the $10 lights. Maybe the under $10 lights are the best, because with hollow pills, I would be forced to plug in a copper shelf anyway....
So frustrating... I'm hearing it again from vendors/manufacturers that the parts are made well enough to handle the stock demands and expectations... Ugh.
The higher the heat load that has to be shifted and the smaller the area that heat has to be transferred through, the more improvement you will see with better thermal paste.
Can you measure a difference in light output under any conditions when testing an excellent paste against a crappy paste, assuming the same prep work?
Me? I got AS5, Fujik, GC Extreme for example but not worth it for me - no time. Really needs hours/days to see how they perform after a few cycles of cells for example. Some will hold up better. You can't just do a fresh test, though it's worth something. I thought someone else did and publshed this all before? The toothpaste test? Thought it was all for naught? But I don't think they included coollab - dunno where the thread/results are anymore...
Even if I were to do it, I'd need a perfectly flat pill, I mean perfect... Not sure I can guarantee that on anything I have.
But we're talking about, what, a max of 35 watts for one XML2 before the bond wires fail? And usually at that level it'll be on a 20mm board? That's a huge area and not very much heat, compared to a test designed to replicate a CPU heatsink scenario. The differences between a good paste and a not-good paste will be small enough at our power/heat levels that they will not show up in the amount of light output. Is that not what matters?
In the CPU world, back when CPUs had a tiny exposed die, quality of the paste and the clamp loads present were ultra-critical. That's when most of the thermal paste shootout comparisons were done. Currently, with nearly everything having a bonded-on heatspreader on top though, the performance of the paste is much less critical, the critical joint has moved inside the CPU between the die and the heatspreader where (most times) you can't get at it, and Intel/AMD use very good quality pastes inside there. Some heatspreaders are even soldered to the die.
And in the other thread where I asked about taking a star-pill contact print when using screws, the reason is that there are only two very small areas around the screws where the clamp load is high enough to transfer dye from one piece to the other. It's pretty horrible, really. And thermal pastes only work when the clamp loads are high enough to transfer the dye, anywhere else the layer is too thick to transfer any meaningful amount of heat, and the paste just acts as an insulator instead. And yet... it still works. There's very little if any difference between using screws where there is only a tiny contact area due to MCPCB distortion, and soldered where there is 100% contact. At the levels possible in a flashlight these minutiae just do not make the slightest bit of difference. If they did, if the difference between a good paste and a shitty paste actually showed up at flashlight power/heat levels, I'd be all for it. I'm not for stuff that doesn't make any difference.