Luxeon LED headlight assembly

Have an old processor heat sink laying around? This Pentium 3 heatsink is the perfect size for four Luxeon I Star/O assemblies (LumiLEDs Product# LXHL-NWE8). Luxeon light sources need very good heat dissipation, and this might be over-doing it, but why not prolong the life of the light sources?

I wanted to get around paying up to $300 for a battery-powered bicycle headlamp, so I decided to meet or exceed the expectations with available parts.

I initially bought only two of these Luxeon Stars (with optics) over a year ago. I didn't have a good way to drive them other than my Elenco Precision (TM) power supply and alkaline batteries, both of which are voltage sources. I tried using two D cell alkaline batteries, this provided 3.0V which is less than the typical 3.4V, and not didn't force enough current to for the LED's to be all the way on. I also didn't have a good mounting bracket, or a good way to mount the batteries to the bicycle frame.


Recently, using the mounting bracket from and old mechanical speedometer and zip ties, I was able to make an excellent mount. This stops the lights from changing pitch (very important for this application) or breaking free. I connected them all in parallel to try to give them all 3.0V from the two D cells.

Lesson learned: Don't solder when the device is a metal core PCB and it's sitting on a heat sink with silicon paste in between! Leave these PCB's on an insulator while soldering.


I noted that they were only drawing around 800mA, nowhere near 1.4A (4x350mA). I decided to increase my power supply voltage potential, but this meant there needs to be another component in series with the light sources to allow the excess voltage to be dissipated. In the application notes from LumiLEDs, they recommend using a 6V power supply and a 10 Ohm resistor for a Luxeon I. I added another battery pack to the frame to get 6.3V, but I decided to use 0.9V rectifying diodes (3A max) rather than four resistors (one in series with each supply) which would require desoldering. This achieved 1.6A with one diode in series and 1.1A with two diodes in series. I stuck with the two diodes in series to stay on the safe side. The diodes get rather hot very quickly since they're dissipating so much power. This is a major problem, by my measurements -- 45% of the power was being dissipated in the diodes themselves, delivering only 55% of the power from the batteries to the light sources. Another reason that a passive approach (linear mode voltage regulation) is bad for alkaline battery applications is that the higher the power delivery, the more the batteries self heat and increase internal resistance. In other words, Alkaline's are less efficient when discharged quickly. It is a common trend that batteries are more efficient with intermittent loading for reasons related to battery theory. For much more, see the IEEE paper Battery­Driven System Design: A New Frontier in Low Power Design, by Kanishka Lahiri, Anand Raghunathan, Sujit Dey, and Debashis Panigrahi.

Lesson learned: The voltage drop across an LED is related to the energy released in each photon--relating to the wavelength of the light emitted. (Note that this is the P-N junction voltage drop, and differs from the external voltage drop and light spectrum is also highly dependent upon construction materials). The current through the LED corresponds to the number of photons emitted--relating to the intensity of light emitted. These brief rules of thumb really help in understanding how to apply high-power LED's.

Lesson learned: Luxeon LED's are constant-current devices, so to get them to the proper luminous intensity and power consumption, they need to be driven at the proper current amperage. For Luxeon I's this is 350mA, for Luxeon III's this is 700mA, etc. This point remained subtle to me because I figured I could always just find a voltage potential which would result in the proper current being drawn. The problem is that LumiLEDs has a tighter manufacturer tolerance of 350mA operating conditions than the forward voltage (Vf) which is loosely maintained. This means each light source will require a slightly different Vf to draw 350mA and use run at the full 1 Watt. The most robust solution is a constant current source which can be used to drive multiple light source in series. This also allows the supply voltage to fluctuate (eg: a draining battery) and not affect the luminous intensity as it would without a regulator.

Lesson learned: A common optimization for alkaline battery-powered applications is to use a charge-pumping power regulator. These are capable of producing voltages higher than their input voltage (12VDC from 6VDC for instance) by intermittently presenting a larger load and driving an oscillating circuit. The typical result is luminous intensity remaining mostly flat until the battery is mostly depleted. Just connecting a 12VDC battery supply would result in an exponentially decaying luminous intensity.

These conclusions lead me to replace my diode-dissipating circuit with a switch-mode power supply including a boost circuit. Because of the scarce availability of through-hole mounting switch-mode power supplies and the necessity for a PCB, I was very happy to find products from LED Dynamics including the BoostPuck. Their Application Figure 3 in their datasheet is my exact application. The regulators are unfortunately around $34, but it's worth the efficiency of a switch-mode power supply, the optimized battery utilization, and proper driving (constant current of 350mA) of the light sources to get the most light without risking burning out the LED.

When the unit arrives, I'll removed the hot glue (to keep the parts clean) and connect the light sources in series and connect the BoostPuck to the ligth sources after the rocker switch. As per the application notes, I'll need to include a capacitor since my batteries are a few feet away. More as soon as the part comes in.