make a 20mm board with two 1oz copper planes and one with 0.35mm plated holes in a 0.55x0.48mm grit
put a 2512 3W resistor in the middle
clamp the outer 1.5mm bare copper ring between 2 aluminum plates that leave the inner PCB open or simply use a flashlight with retaining ring
measure the resistors temperature of both boards after 10 minute when its saturated
then talk again about sense and senseless of thermal viases
.
you forget that a via is a plated copper hole, so there is a lot copper inside the board also conducting heat radial, the thicker the board the more copper you add
even if you remove FR4 the copper conducts a lot more heat than the removed material
and you forget the main cooling comes from the bare copper rim of the driver touching the metal body,
getting the heat there is our goal and viases do it better
surface heat conduction to air is totally irrelevant inside a flashlight
you talk about removing half the surface copper this is totally right!
0.35mm plated hole removes 0.28mm copper circle in a 0.55x 0.48mm grit
so 0.12mm² copper removed inside a 0.264mm² area
but a via 0.35mm on a 1.6mm thick board adds about 5 times the copper you removed on surface inside the driver
calculate the copper inside a board
we remove 0.35mmx1.6mm FR4 this equals 0.31mm³ per via
FR4 thermal conductivity 0.25
0.25*0.31mm³=0.08
solid surface copper board volume
2*2π0.175²*0.035mm=0.0135mm³
Kupfer thermal conductivity 9355
0.0135mm²*9355=126 so the copper planes in 1 oz conduct more than 1000 times the heat than the FR4 does
so we can simply ignore the FR4 at all in the thermal equasion
surface copper removed 0.35mm hole minus 2 times 35um copper
A=2*2π*(0,175mm-0.035mm)=0.25mm²
V=0.25mm²*0.035mm=0.0086mm³ 0.0086mm³*9355=80,5
lets look at the internal copper added
internal copper added A=(0.35-0.035)*0.035*π=0.034mm²
V=0.034mm²*1.6mm=0.055mm³
lets calculate the copper difference between the added and removed copper
0.055mm³-0.0086mm³=0.046mm³
thermal conductivity gained after calculating lost surface copper and added internal copper 0.046mm³*9355=430
so in the end the via copper conducts 430 while the removed copper would conduct 80.5 this is more than 5 times more heat conduction based on the coppers volume, sure the hollow cylinder does not conduct as good as a solid plane but in the end its likely still 4 times better
Its nice you theorize about the viases, practically they work like 2 times better than without them
As said, conduction of heat to flashlight body can sometime be not as good as you imagine because powerful light (which require all those via anyway) likely has a very bright led, which make a lot of heat, so the local temperature near the ring around driver PCB can in some cases be quite warm, and thermal cooling by conduction may not be as good as you imagine. In event of efficient switching circuit on driver (versus linear), it is possible (at high led power) and high air temperature that the driver is cooler than the case.
As you mention, the bulk of copper is in the form of via, not in-plane copper. Adding holes to the in-plane copper significantly reduce in-plane thermal conduction (at the extent of the hole per area ratio). Therefore thermal conduction is not as good as you imagine. Immediate thermal sinking due to increased copper volume will help initially, but once thermal saturation occur, transmission is not as good as you imagine.
Overall I think adding so many via is necessary and not provide much benefit (which is fine), but this come at a cost of significantly longer fabrication time and structural rigidity and board flex (lead to component solder joint breaking), which I think is not a good trade-off to make for robust and strong flashlight.
It's not simple, there is always an optimum number and size of vias for such designs; with too many vias close to heat source you are creating bottleneck and cutting "heat supply" via top/bottom copper layer for more distant vias, so they become useless.
not simple; ‘more copper’ is not always better, because where the copper is, is important. Many scenario:
- if the goal is to sink heat away from source by a good cooling source like heatsink, radiator etc, then having additional copper mass is not ideal. instead, you need to have lowest thermal resistance away from heatsource. one example is thermal pipe used in cpu, instead of big copper block.
- in the above pcb where so many vias a drilled, then the surface copper is significantly reduced creating poor thermal path in-plane. yes you get better thermal conductivity plane-to-plane in z-axis, but too many via will reduce effectiveness for in-plane conduction.
- as a result, to conduct heat away from source, you can put many via near source to conduct heat from top to bottom plane (or internal if it exist), then you want to then reduce via away from source to maximize in-plane thermal conduction to the heat sink (flashlight body)
- in addition, adding so many via will lead to significant structural integrity tradeoff, which i think is no want for flashlight where the middle of the pcb contact battery
however for fun and to look interesting, this design is ok. i just want to suggest improvement in term of engineering practice. thank you for consideration!
Okay, point taken.
But you still make the fibre glass PCB conduct heat better, ALSO TO THE OUTER PERIMETER, by drilling out the fibre glass and inserting copper vias.
Lexel explained this too.
He even tested it.
there are very reliable things to improve it, but they can easily escalate the price of the board
1. add viases —> at our quantities for free
2. 2oz copper —> relative cheap but needs thicker traces and solder lands
3. 4 layer —> in chineese fabs increase costs quite a bit
4. ceramic PCB replacing fibre glass —> expensive
5. special thermal copper layer like 0.3mm inside the board —> very expensive, needs big batch to be produced
Not for sale yet. In previous tests he said it would do 6A output. At lower battery voltages the amp draw from the battery could get as high as 15A-16A IIRC, so it can put out some real power. Of course, the output level can be turned down to what you want. I think it’s a resistor swap to limit output. We just have to be patient.
BTW, I’ve been waiting about a year, so do not hold your breath. Lol
Is it possible to build it myself with your pcb and list of parts and program code? Small 6A boost driver is very useful for me. Thank you lexel. It seem some people like you, loneoceans, schoki is making such a driver, but so far i only see loneoceans with working design and flashlight. I am hope to build one myself.
Mtn e is currently making and selling a boost driver, but it is only 17mm and for clicky lights at the moment. Bigger sizes and e-switch versions will be sold soon.
yes a lot depends on the used Bost chip and thermal cooling, on my driver I would strongly suggest to pot around the Boost chip with aluminum oxide glue
you could try the same on MTN driver
Basically whats limiting output current atm is thermal properties of the board
4.2A? That must be due to its small 17mm size (along with the 1.3A 12v output). The bigger ones are 6A just like Lexels (12v also goes up to 2.5A-3A). Roughly 50W seems to be the most you can get from these smallish sized boost drivers.
Lexel, can your 24mm drivers be sanded down to 23mm?
I have the Amutorch VG10 that takes a 23mm driver, and I guess one of yours would be good to fit in, if I can sand it to 23mm! BTW, will it work fine with:
a) Forward Clicky Switch
b) OSRAM KW CSLNM1.TG – 1mm2
Thanks in advance!
BTW, about the driver for the BlitzWolf BW-ET1, I will let it stay still until I get a good solution! Thanks for your answers, though
