Trying to design a bike light
November 12, 2006
I’ve been reading a lot about LEDs, power supplies, components, and the response of the human eye to light. For my purposes, the following things seem to be true about LEDs:
I am stuck on the power supply issue. It seems to be a standard that the lights should come on at 5.5mph. I need to calibrate the power output of my little generators to see how much they produce at that speed. My old generators are clearly unregulated; I cooked a few headlights coming down hills years ago. LEDs are much less tolerant of over-voltage. So I need a regulator. There are several designs that I am considering.
One is a straight current-source regulator, out of EE textbooks, probably built using a power MOSFET so that I can make the turn-on voltage as low as possible. This has the disadvantage of pulling a constant current from the generator, no matter what the voltage. At high speeds, I would not only be dumping some number of watts into the MOSFET, I would be taxing my own self to do it.
A second choice is a switching power supply. The National Semiconductor LM3478 looks like a good choice, and they have lots of design notes at their site. I’m particularly interested in the SEPIC designs, because they can take a high or low input voltage and regulate it to a constant output. There are four sources of power loss here. One is in the switching diode, which will consume 0.5 volts. The second is in the current sense resistor, which will probably consume 0.1 volts. Switching losses in the MOSFET are at high frequencies. Resistive losses in the power supply during the charging duty cycle also contribute.
A third choice is a cheesy variant on a switching power supply. This would simply use the input voltage to charge a capacitor, and use a MOSFET to gate current into the capacitor, based on current feedback. The inefficiencies here are resistive losses in the MOSFET, resistive losses in the current sense resistor, and resistive losses in the MOSFET during switching.
A fourth choice is not to conserve excess power, but instead to dump it into additional lights. When riding fast, I want as much light as possible, and that could well be more than the light supplied at lower speeds. One easy way to do this to set up multiple simple-switching power supplies for larger and larger stacks of LEDs, run in parallel, and to gate the MOSFETS from the overall input voltage.
One thing I did learn — Schottky diodes are probably my friend. I can save 0.2 V, or 28%, on my forward voltage drop over silicon (if I could find a power Ge diode, that might be even better).
The fourth choice is looking pretty good, except that my transistor design skills are stale, and my FET design skills are really stale.