modamag
Flashlight Enthusiast
It's not a flashlight but have everything to do with almost everyone of us.
I promised some people a while back some thermal modeling of our beloved Mag mods, well here it is.
Case #1 - PSS (Modded Hotlips/Osink) - Single Luxeon III
Nothing special here
Case #2 - PSS (Modded Hotlips/Osink) - Hi-Output Single Luxeon V
Worst case scenario for a single Luxeon, LuxV @ 1.2A (~8.5W) ... well for normal people. I don't count unnerve's LuxIII @ 3A the typical case now.
Case #3 - PTS-D - 3x LuxIII @ 1.0A each (~10W Total)
This could handle the 10W power source but it's pretty much at its limit.
The same would apply to any "disk" profile heatsink. It's just not enough height to effectively transfer the heat out radially.
Case #4 - PTS2-D - 3x LuxIII @ 1.0A each (~10W Total)
10W it's not much of a problem. One of the key feature, E-Can, actually have a major negative effect on the heat transfer. If the E-Can is not present the thermal transition will be much smoother and w/o a localize hot spot in the rear.
Case #5 - PTS2-D - 3x LuxIII @ 1.0A each (~10W Total) with standard Mag head
Notice how all the heat is rushing to the rear.
Case #6 - PTS2-D - 3x LuxIII @ 1.0A each (~10W Total) with grooved Mag head
Surprise! So am I, let me attempt to explain. This model deals with stagnant air where the heat transfer coefficient for natural convection is only 1-50 W/(m^2K), in our case 5 W/(m^2K) While force convection will be dozens of times higher. When conduction (hand) is greater than convection (air) adding fins is a waste. It only create localize hot spot. However, if there is some air flow thru these fins it greatly help with the cooling.
Case #7 - PTS2-D - 3x LuxV @ 1.0A each (~21W Total) with grooved Mag head
Insane case for those who wish to drive 3x LuxV @ 1.0A (~21W). You better put a dimming control circuit in that thing.
I'm not even gonna attempt the 4xLuxV scenario.
Case #8 - PTS2-D - 3x LuxIII @ 10A each (simulating ~100W hotwire mod) with grooved Mag head
Since most sources indicates that 90% of halogen/incandescent light are infrared so only 90W of power is applied to the model. Forget about burnt marshmallow; watch your roast skins instead.
* Pay attention to the temperature scale; don't just look at the beautiful colors now.
* Model values are 10-20F less than the actual values that I've seen in real time test w/o hand contact as a sink.
* Case 1-4 does not utilize hand contact as a sink temp, where case 5-8 does.
=== Assumptions ===
- Maximum hand contact surface temp = 110F / 43.4C
Biggest assumption! No matter what's the light source, you 97.8F hands will attempt to maintain 110F.
- Edge Stablelization temp = 120F / 48.9C
2nd biggest assumption. This is valid for 7-9W w/o hand contact. You actually have to perform a runtime test to get this number FOR EACH TEST CASE.
- Ambient stale air temp = 70F / 21.1C
- Natural Air convection coefficient = 5 W/(m^2K)
- All mating surface are flushed/merged
- Does not account for radiation or convection in pockets
- Ignored all heat generated by the batteries system
* If there is any major flaws in these assumptions please let me know and I'll update the model.
=== Conclusion ===
It's amazing the affect of liquid cooling (your burnt hands) on the system. It's much more effective of a cooling mechanism than the air cooling.
Realistically you can put a 500W power plant in a Maglite but as long as it's for "short" duration then it's relatively safe. Great examples of these are bwaites' USL & cmac's "the Torch". This is because it takes some time for the heat to saturate all the components (reflector, batteries, switch, etc).
It's when modders pushing the limit to increase the overall output AND runtime that's when it's a problem. Practicality and safety are sometime placed in the back seat.
I hope this information can help future generations of Mag modders.
=== Side Note ===
It's amazing how much PC have progressed over the years. Back in the days @ RPI, it took our IBM RISC station (latest & greatest @ the time) over 40 minutes to generate the results with coarse meshing on ProE (or was it UniGraphics :thinking. Nowadays, my 16m old PC can do ultra fine mesh within 10 minutes
I promised some people a while back some thermal modeling of our beloved Mag mods, well here it is.
Case #1 - PSS (Modded Hotlips/Osink) - Single Luxeon III
Nothing special here
Case #2 - PSS (Modded Hotlips/Osink) - Hi-Output Single Luxeon V
Worst case scenario for a single Luxeon, LuxV @ 1.2A (~8.5W) ... well for normal people. I don't count unnerve's LuxIII @ 3A the typical case now.
Case #3 - PTS-D - 3x LuxIII @ 1.0A each (~10W Total)
This could handle the 10W power source but it's pretty much at its limit.
The same would apply to any "disk" profile heatsink. It's just not enough height to effectively transfer the heat out radially.
Case #4 - PTS2-D - 3x LuxIII @ 1.0A each (~10W Total)
10W it's not much of a problem. One of the key feature, E-Can, actually have a major negative effect on the heat transfer. If the E-Can is not present the thermal transition will be much smoother and w/o a localize hot spot in the rear.
Case #5 - PTS2-D - 3x LuxIII @ 1.0A each (~10W Total) with standard Mag head
Notice how all the heat is rushing to the rear.
Case #6 - PTS2-D - 3x LuxIII @ 1.0A each (~10W Total) with grooved Mag head
Surprise! So am I, let me attempt to explain. This model deals with stagnant air where the heat transfer coefficient for natural convection is only 1-50 W/(m^2K), in our case 5 W/(m^2K) While force convection will be dozens of times higher. When conduction (hand) is greater than convection (air) adding fins is a waste. It only create localize hot spot. However, if there is some air flow thru these fins it greatly help with the cooling.
Case #7 - PTS2-D - 3x LuxV @ 1.0A each (~21W Total) with grooved Mag head
Insane case for those who wish to drive 3x LuxV @ 1.0A (~21W). You better put a dimming control circuit in that thing.
I'm not even gonna attempt the 4xLuxV scenario.
Case #8 - PTS2-D - 3x LuxIII @ 10A each (simulating ~100W hotwire mod) with grooved Mag head
Since most sources indicates that 90% of halogen/incandescent light are infrared so only 90W of power is applied to the model. Forget about burnt marshmallow; watch your roast skins instead.
* Pay attention to the temperature scale; don't just look at the beautiful colors now.
* Model values are 10-20F less than the actual values that I've seen in real time test w/o hand contact as a sink.
* Case 1-4 does not utilize hand contact as a sink temp, where case 5-8 does.
=== Assumptions ===
- Maximum hand contact surface temp = 110F / 43.4C
Biggest assumption! No matter what's the light source, you 97.8F hands will attempt to maintain 110F.
- Edge Stablelization temp = 120F / 48.9C
2nd biggest assumption. This is valid for 7-9W w/o hand contact. You actually have to perform a runtime test to get this number FOR EACH TEST CASE.
- Ambient stale air temp = 70F / 21.1C
- Natural Air convection coefficient = 5 W/(m^2K)
- All mating surface are flushed/merged
- Does not account for radiation or convection in pockets
- Ignored all heat generated by the batteries system
* If there is any major flaws in these assumptions please let me know and I'll update the model.
=== Conclusion ===
It's amazing the affect of liquid cooling (your burnt hands) on the system. It's much more effective of a cooling mechanism than the air cooling.
Realistically you can put a 500W power plant in a Maglite but as long as it's for "short" duration then it's relatively safe. Great examples of these are bwaites' USL & cmac's "the Torch". This is because it takes some time for the heat to saturate all the components (reflector, batteries, switch, etc).
It's when modders pushing the limit to increase the overall output AND runtime that's when it's a problem. Practicality and safety are sometime placed in the back seat.
I hope this information can help future generations of Mag modders.
=== Side Note ===
It's amazing how much PC have progressed over the years. Back in the days @ RPI, it took our IBM RISC station (latest & greatest @ the time) over 40 minutes to generate the results with coarse meshing on ProE (or was it UniGraphics :thinking. Nowadays, my 16m old PC can do ultra fine mesh within 10 minutes