Since another has opened the door on design input, I also seek counsel.
(Mad! Completely nuts!, No NOT that sort of counsel, or a lot of us would be in trouble, right?)
You can skip to the bottom to the issue and refer to the following sections to understand why I am asking, or where I messed up.
Background on why I want to use copper end caps at all (FYI):
I would prefer a CNC aluminum head light unit for the minimal number of thermal interfaces and the 'look' but those are beyond my budget at the moment. I also wish to try a unique arrangement of XPG's to double up a usable light and an experimental platform for road beam testing. I want the finished lights to look like a commercial setup and be amenable to redesign later. Lastly, in case one or a few readers wished to follow the same path, I thought at least a limited supply of reasonably priced bits would be good to base these lights on.
So I decided to try two of EL34's stash of MR11 (not MR16 as listed) bullet Marwi metal halogen light heads, and an array of matching handlebar and helmet mounts to add mounting positions to the mix of testing. He also has the older funnel or bell shaped head units and a reflector/heatsink machined to fit a P7, and you can piece out a P7 light kit as you like, as some of you know. However, I wanted a multi XP-G design with the new LISA2 pin lenses. I wanted room for either a Bflex or Maxfflex driver board behind the light engine array so the bullet style is it.
That leads to the reason for considering the copper pipe end cap as the starting point for the MCPCB mounting and heat spreader to the aluminum shell.
For those interested, Google this forum for copper and aluminum and read away. The summary is that copper is as much better at conducting heat from hotter to cooler zones compared to aluminum as aluminum is compared to steel. Silver is a bit better than copper but no where near the cost differntial. Since we are concerned with long term steady state and not short-term transient heat storage as in some flashlight applications, the higher heat storage per unit volume of copper is not particulary useful.
Copper is more expensive and heavier than aluminum. But pipe caps are mass produced and you can buy just what you need and not 4' or aluminum tube when you need 50 cm, so at the DIY level of cost, copper can be cheaper. The weight can be a deal breaker for some cycling applications. For non-competition or training use, as long as we aren't talking a solid copper CNC light, a copper heatsink is a worthy alternative to aluminum.
And so, we arrive at the use of copper pipe end caps. It so happens, that a 1" pipe end cap is a good first approximation of a nesting heat sink for the Marwi light with a wall to mount the led and flanges to move the heat to the body of the light and at least one builder has used this method successfully it in a triple Q5 application with some cut away to allow for the mounting slot.
Go to post 24.
LED thermal management 101 (background for issue):
However, if you Google this forum on lapping heat sinks or stars, there is an issue with the end cap and MCPCB mounting that needs to be addressed for optimal thermal path. Although the thermal path from the mounted LED to the MCPCB is the most critical, I have elected to trust Cutter for my mounted XP-G's (as so many here have, so they are what they are, but I hope there are some VIAs from the thermal pad to the bottom surface). That leaves me with the MCPCB to heat sink and heat sink to body interfaces to optimize. It is the first interface that isn't as simple as it first appears.
The MCPCB is a heat spreader and metal-metal contact between the spreader and heatsink is an order of magnitude better than thermal glues or greases, which are orders of magnitude better than an air gap. So you want as much metal to metal contact as possible, the thinnest layer of thermal compound or epoxy, and no air gaps.
MCPCB's are punched or cut and by definition are not flat. Some DIY's have lapped these while being careful not to rock them and flare the edges. (if anyone has lapped the 10 mm round ones and has a good technique, now's the time to come forward).
Even new heat sinks may not be ruler flat. Even if flat, unless they are carefully polished, they have microscopically large peaks and valleys (polished equals microscopicaly smaller peaks and valleys). Thermal compounds and epoxies fill these gaps and fill the air gaps between the heat sink and MCPCB. But if the layer is thick in order to compensate for a considerable lack of flatness, then the thermal resistance at this interface will be high and can prevent heat leaving the MCPCB as fast as needed to keep the LED in operating range let alone as close to ambient as possible.
Finally: The copper end cap issue:
This is true for all heat sinks, so why am I reviewing all this for the copper end cap? First, for anyone familiar with these, the end wall or floor of the cap is far from being planar, let alone polished. Second, is the teacher in me: I spent a lot of time tracking all this down and thought it would be nice for onter newbies and not so newbies to know. Last, I need any misunderstanding to be corrected and there are willing parties here.
I have two manufacturers' examples, and though this is a small non-random sampling, 100% is a high frequency, so it may be a common problem. I have two different logos stamped in the copper and ghostly images of them clearly visible inside the cap bottom. Copper is maleable and ductile which means it can be molded and bent to shape, handy to modify the end cap to fit the inside of a bullet-shaped light body. It also means the pressing has pushed up the center bottoms of these caps. I can detect the raised parts around these logos inside the pipe caps and the recessed circumference near the side wall. Not even close to flat, let alone polished.
So has anyone addressed this issue to get a flat mount for the MCPCB?
The XP-G has a high thermal limit and a slower heat decay, so is this is a lot of fuss over a slight reduction if lumens/watt?
Does this solution sound feasible:
Unless I learn better here, I plan to cut off the bottom of one 1" end cap, remove any 'edge' of the former 'wall' remaining, then sand the inner side of the bottom flat and lap it to a mirror finish. Then, lap the MCPCB's. Mark, drill , and tap mounting holes for screws, and use minimal thermal paste on the MCPCBs.
As it happens, I need a slightly larger diameter than a 1" pipe end cap to meet the inner wall of the body from just behind the former seating for the M11 bulb to the plane of the LED and lens mounting and a little beyond. So I plan to use that the removed wall, (which is a short cylinder after the bottom is cut off). It will be cut lengthwise to make a split cylinder that can be expanded and used to make the heat sink larger in diameter to meet the inside of the body at that point.
I am also mounting the LED's asymetrically, so I need a thicker thermal path to the opposite side of the light. (Cross section of heat sink should be about 1/4-1/2 the distance from the LED to the body, a rule of thumb in this forum, under heat sink size rule of thumb thead, I think) I will use a second pipe cap's bottom for this added thickness and it's walls will be beveled to match and meet the bullet shape towards the rear of the light to help heat flow through that part of the light's body.
Assuming I am a sucker for all the hand work, any thoughts?
I can assemble the heat sink in two ways:
Each light will use about 3.5 watts on the street on 'highs' (350 mA) and 10 on the trails (1 A), about 80% of that, as heat.
I can solder the the heat sink as one assembly and have difficulty mounting the MCPCB's and thinning out the thermal paste for best thermal path because they are close to the sides at the bottom of a 8 mm deep 'well'.
Or I can make the end cap bottom a removeable part of the heat sink bolted and thermal pasted to the second part with lapped mating surfaces, rather than soldered in place. Then the MCPCBs and lens can be mounted in the open to aid thinning out the thermal paste.
In the first case, I get the excellent thermal path of solder through the heat sink but likely a poorer path at the MCPCB's. In the second design, I get easier modification, likely much better MCPCB thermal path, but I add another resistance interface to the thermal path to the body.
In both cases, I plan to hone the heat sink to the body and use thermal paste to reduce the resistance at the last thermal interface before the air. (CNC will start to look good, during this process, I'm sure).
Does anyone have experience or a gut feeling about how the 2-piece heat sink would work compared to the 1-piece?
My gut says the smaller and hotter interface needs the most attention. So the 2-piece is the better solution thermally and for future mods.
What do you think?