DFiorentino
Flashlight Enthusiast
Due to the fact that I have several mods I'm getting ready to work on, I wanted to sort out my LED supplies based on performance. Using the bin code just wasn't enough for me, so I home brewed myself a budget "test box". It is simply a 1/2 gallon milk carton with the exterior painted metallic silver then covered in electrical tape to prevent light infiltration. I have my eBay light meter attached to the "bottom" (in use this becomes the side) again covered in electrical tape to affix it and prevent light from entering. On one "side" (in use this becomes the bottom) I have cut a hole that fits around a 2" diameter stepped and finned CPU heatsink. I tested only Luxeon stars and they were stuck to the heatsink using Ceramatique and wires were soldered to provide the electrical connection. I powered the stars with a bench power supply and monitored the voltage and current with two Fluke DMMs (don't have the model numbers handy). It's too embarassing to take pics of, but suffice to say that it's good enough to compare LEDs I have at hand. All of my light readings were taken in LUX. I tested at 350mA, 500mA, 700mA, and 1000mA. And here are the results:
BLC = Bogus Lumen Conversion
Seperated stars and emitters:
__________________________________________________
Stars LUX graph:
Emmiters LUX graph:
Mixed 1w & 3w LUX graph:
This sort of helps explain the LuxI/LuxIII battle. For the most part it seems R-bin LuxIs and T-bin LuxIIIs are comparable as well as S-bin LuxIs and U-bin LuxIIIs. However, it is more likely that the underdriven LuxIII will suffer from a shift in tint when driven at below spec power levels. Food for thought.
5w LUX graph:
__________________________________________________
Stars Vf graph:
Emmiters Vf graph:
__________________________________________________
Stars efficiency (LUX / Watt) graph:
Emmiters efficiency (LUX / Watt) graph:
__________________________________________________
Temperature Data:
Temperature readings were done by affixing the emmiter to a D size O-sink via Ceramatique. Temperature was taken in Celcius via contact probe at the raised platform for the emmiter to get the closest junction temperature possible. The O-sink was placed on a 2" CPU heatsink to simulate being installed in an actual flashlight. This last part was critical as I found out. I tested the luxV on the O-sink alone out of curiousity and saw temps of 110 degrees Celcius (230 deg.F) after 5 minutes. (Don't even think of using thermal paste at this high of a temperature. It basically turns into a liquid. Use Thermal epoxy at this point.) I chose to test at 1 minute as it seemed to be a maximum on time for momentary type action. I also tested at 5 minutes as it seemed the O-sink stabilzed within this time period.
Calculated junction temperatures were based on actual Vf and current inputs. Thermal resistance values taken from Lumileds documentation for emmiters.
5 Minute Graph (Calculated Junction Temperature; out of Mag):
The O-sink in a Mag is a great heatsink. In the last set of test data, with the O-sink installed in the Mag, the light was basically in candle mode. With a hand holding the light, things would be marginally better. Looks like 1500mA is the bleeding edge for the luxI and luxV. For a show light or as a burst mode its feasible, but for sustained use, I'd keep things at 1300mA or below. And that is with this optimally heatsinked test setup. Looks Like the luxIII was still fairly stable though, even at 1500mA.
-DF
Disclaimer: I have conducted this testing on my own behalf for my own benefit. If others find this useful, that is great. However, please take into account that this is a comparative analysis and not absolute findings. YMMV
BLC = Bogus Lumen Conversion
Seperated stars and emitters:
__________________________________________________
Stars LUX graph:
Emmiters LUX graph:
Mixed 1w & 3w LUX graph:
This sort of helps explain the LuxI/LuxIII battle. For the most part it seems R-bin LuxIs and T-bin LuxIIIs are comparable as well as S-bin LuxIs and U-bin LuxIIIs. However, it is more likely that the underdriven LuxIII will suffer from a shift in tint when driven at below spec power levels. Food for thought.
5w LUX graph:
__________________________________________________
Stars Vf graph:
Emmiters Vf graph:
__________________________________________________
Stars efficiency (LUX / Watt) graph:
Emmiters efficiency (LUX / Watt) graph:
__________________________________________________
Temperature Data:
Temperature readings were done by affixing the emmiter to a D size O-sink via Ceramatique. Temperature was taken in Celcius via contact probe at the raised platform for the emmiter to get the closest junction temperature possible. The O-sink was placed on a 2" CPU heatsink to simulate being installed in an actual flashlight. This last part was critical as I found out. I tested the luxV on the O-sink alone out of curiousity and saw temps of 110 degrees Celcius (230 deg.F) after 5 minutes. (Don't even think of using thermal paste at this high of a temperature. It basically turns into a liquid. Use Thermal epoxy at this point.) I chose to test at 1 minute as it seemed to be a maximum on time for momentary type action. I also tested at 5 minutes as it seemed the O-sink stabilzed within this time period.
Calculated junction temperatures were based on actual Vf and current inputs. Thermal resistance values taken from Lumileds documentation for emmiters.
5 Minute Graph (Calculated Junction Temperature; out of Mag):
The O-sink in a Mag is a great heatsink. In the last set of test data, with the O-sink installed in the Mag, the light was basically in candle mode. With a hand holding the light, things would be marginally better. Looks like 1500mA is the bleeding edge for the luxI and luxV. For a show light or as a burst mode its feasible, but for sustained use, I'd keep things at 1300mA or below. And that is with this optimally heatsinked test setup. Looks Like the luxIII was still fairly stable though, even at 1500mA.
-DF
Disclaimer: I have conducted this testing on my own behalf for my own benefit. If others find this useful, that is great. However, please take into account that this is a comparative analysis and not absolute findings. YMMV
Last edited: