Finally I did a proper torture test of LD-2 with external MOSFET.
Since LD-1 beginning some members wanted to use this tiny driver(designed for small/med size high power lights) for driving big setups,from XM-L2 or XPL triples,and these days triple XHP50s or 70s,or MTG2s isn't unusual.So we're talking about potentially hundreds of watts of input/output power,and in extreme cases, 10+Watts of power dissipated in driver,and that's a lot of heat for common "float in the air" drivers; this is amount of heat generated by XM-L2 at 3+A!
"Expander boards" (under LD-1 info thread) were potential solution;idea was to put MOSFET(main source of heat in linear driver) and other necessary components on separate PCB with good thermal path to flashlight body,so that heat from mosfet could be dissipated more easily.In reality "expander boards" did offer higher heat dissipation,but they were too complicated to use/connect,and since bottom side was connected to mosfet's drain,additional insulating layer between board and flashlight was necessary,overall just too complicated.
After LD-2 announcement,again similar questions about high power setups appeared.
Only choice was again some soft of "external" mosfet mounting,but this time for LD-2 only mosfet needs to be on external PCB,which simplifies things a lot. I didn't like FR-4 vith vias solution on expander boards;they don't have great conductivity, and as I mentioned they must be further insulated from flashlight with kapton/mica sheet+themal grease/paste.
Another solution is custom designed MCPCB(metal core pcb),just like for leds,but I figured out,I can use existing XM-L MCPCBs with a little modification (remove copper trace under Gate pin):
Only three wires are needed(two thick wires and one thin wire for gate) for complete connection,and MCPCB bottom side is already electrically insulated so this pcb can be mounted/glued directly to heatsink/flashlight body,just like typical led star mounting.
I used XP-G2 triple (parallel) for my torture test(s),mosfet board was glued with Arctic Alumina adhesive on test heatsink. LD-2 is in old convoy C8 pill for testing purpose. Thermal probe+UT-71C is used for mosfet case temperature monitoring.
Wiring is simple: mosfet's drain(red wire) is connected to leds cathode(-), mosfet source(black wire) is connected to mosfet source on LD-2 board,mosfet gate(thin black wire) is also connected to mosfet gate on LD-2 board,driver GND remains connected to Battery GND,leds anode are connected directly to Batt + via thick wire,driver Vcc-spring pad is connected via thin wire to Batt +.
First test: XP-G2 x3,1xTF32650,1s setup,6A high
Since dissipation on high is only~1W no problems there,even stock driver without cooling can handle that.
Second test: XHP70,2xTF32650,2s setup,6A high
Voltage drop over mosfet: 1.25V
Power dissipated in Mosfet: ~7.5W
Mosfet case temperature: ~90C
Heatsink temperature: ~30C
Calculated thermal resistance mosfet-heatsink: ~8C/W
~8C/W was higher than expected,that means mcpcb has around 6.5C/W thermal resistance,which is little bit higher than expected(but those are cheap Chinese mcpcbs,so no surprise there).
Max. allowed mosfet temperature according to datasheet is 175C,so 7.5Watts is no problem.
Third test:XP-G2 x3,2x18650,2s setup,6A high
In third test parallel xp-g2 x3 is driven from two partially discharged LG D1 in series(yes,1s leds to 2s cells),this way I could easily generate much more heat in mosfet,since voltage drop is very high (~2.5V).
Voltage drop over mosfet: 2.5V
Power dissipated in Mosfet: ~15W
Mosfet case temperature: ~135C
Heatsink temperature: ~30C
Calculated thermal resistance mosfet-heatsink: ~7C/W
Similar result like in 2nd test,mosfet was preety hot at 15Watts of dissipation,and this was with heatsink at 30C.
In flashlight max. case temperature would be ~60C,so 15Watts would cause ~165C mosfet temperature,I wouldn't go any more that that,so conclusion is that maximum allowed power dissipation with mosfet mounted on cree mcpcb is ~15W. This should cover 95% of builds.
I expect results to be better with branded mcpcbs,like Berquist,because they use thinner dielectric with a little bit higher thermal conductivity:
http://www.digikey.com/product-search/en?FV=fff40008,fff80475,fffc00d3&chp=0
Also tested 20mm version of mcpcb,and results are identical,as expected.
Last test was with mosfet mounted on noctigon;test was identical to 3rd one,except cells were two fresh TF32650:
Voltage drop over mosfet: 4.2V
Power dissipated in Mosfet: ~25W
Mosfet case temperature: ~65C
Heatsink temperature: ~30C
Calculated thermal resistance mosfet-heatsink: ~1.4C/W
Since noctigon is direct thermal path,resistance of mcpcb is really negligible,and thermal resistance is much lower,only 1.4C/W which is basically datasheet number for tested mosfet thermal resistance. Even at 25Watts of dissipation mosfet is only ~35C warmer than heatsink,so theoretically it could take much more.
But in this case noctigon's lack of dielectric layer is disadvantage,because mosfet drain is connected to whole noctigon body,and must be insulated from flashlight.This might be useful for very high power setups,but for most of the builds common al mcpcb will be good enough.
I'll post schematic later.