3 Watt LED Bike Light Experiments
February, 2008, Rev c
Michael Krabach


Synopsis of Experiments
Prototype 1 - Resistor Controlled Prototype
Prototype 2 - Triple Cree with large heat sink
Prototype 3 - Double Cree with Individual regulators
Prototype 4 - Single Cree and regulator
Prototype 5 - Single Cree and regulator
Prototype 6 - Triple Cree revised heat sink
Prototype 7 - Triple SSC in C Mag-lite Head
Prototype 8 - Triple Cree in D Mag-lite Head
Prototype 9 - Prefabricated 3 LED Mag-lite Head
Prototype 10 - Red flasher with Wide Optic Lens
Prototype 11 - Yellow flasher no Optics
Prototype 12 - Red flasher no Optics
Prototype 13 - Red flasher two LEDs
Prototype 14 - Triple Cree with 8 deg Optics
Prototype 15 - Red flasher two Cree LEDs
Prototype 16 - Red flasher two LEDs Clear Case
Prototype 17 - Auto 12 v clearance light
Light and Beam Measurements
Conclusions, Recommendations, and Further Speculations
Summary Table and Parts Sources

Prototype 9 - Prefab Triple LED in C Mag-lite Head

Prototype 8 was considered using DX SKU-5225 a prefabricated plate with three 3 watt LED emitters installed, but I used my own assembly with a 1/8 heat sink slug, With the extra parts I decided to finish that version using a C Mag-lite head. This plate is only 1/16 thick vs the 1/8 thickness for my hand cut slug in Prototype 8, but I wanted to see if it dissipates the heat. The LEDs appear to be Cree but there are no specs to be sure. The construction follows the same pattern as Prototype 8. You will have to drill and tap a hole in the head, being careful where it will go with relation to the clamp. To drill on a round surface, file a small notch then drill a pre hole with a small drill. Then enlarge the hole and tap it. I used x20 cap screw. You will have to saw off the excess length of the cap screw so that there is no interference inside with the LED assembly. I cut the battery barrel down as much as I could and still have room for the Luxdrive 3023 Buckpuck and the 5K ohm pot.

The tri-optic lens I used was the DX SKU-1912 with a clear lens. Since the prefab LED plate has holes that match (another coincidence) the optic's feet, do not cut off the tips of the feet. They are needed for proper alignment of the lens. The feet tips are seen below. Then partially assemble the LED unit for wire lengths and then cut and solder all the wires. Install all the components from the front and check for fit and orientation. It is tight fitting everything in the C Mag-lite size.

You will not be able to quickly swap optics like in Prototype 8. The optic feet fit tightly into the heat sink alignment holes, and the heat sink is stuck to the Mag-lite head with thermal paste. If you do get the optics off without unsticking the heat sink, you will not be able to get the optic's feet back into the alignment holes. So changing the optics requires dismantling the LED assembly, including the pot and connector.

As discussed previously, all Mag-lites are different to some extent. In this case because I wanted to use the prefab triple LED module, I had to use a tri-optic lens that dropped down too far in the head for the bezel to hold everything tight. (I could not use the taller tri-optics used in Prototype 7 because they only work with SSC or Luxeon LEDS.) I used a thin piece of plastic white notebook cover as a spacer between the bezel and the tri-optics. A careful cut will force the LED assembly against the head land. Thermal paste was used on the head land for heat transfer.

The final prototype is only 1 oz less in weight from the Prototype 8 with the D head, 9 oz vs 8 oz. Considering the tightness of the assembly and that a plastic spacer is needed on the C version, I would suggest building the D version. The beam for this prototype is slightly different from the same tri-optic in the prototype 8. It is a slightly broadened 14% beam with a slight yellow tinge on the outer beam. I built this prototype quite fast compared the other lights and the prefabricated 3 LEDs certainly make things much easier, even though the beam quality is not as good as the others. I also question the ability of the design to dissipate heat sufficiently based on the tests with prototype 7. While the unit does not get hot to the touch, I have no way of determining the internal temperature.