This time I tried a little different approach than usual: no decorations around the frame, just one reasonably large red taillight, and the whole remaining power of the generator aimed forward. The result is the most powerful darkness exterminator I have ever built. It has an integrated USB charger again, this time with a better regulator. The system is now used on the Bullet of Bílsko.
The circuit is designed for any 3 W dynamo, sidewall or hub. A dynamo laced in smaller wheel than intended for shouldn't hurt it and provides light even at very low walking speeds, but wastes power in Zener diodes at higher speeds.
40 warm white LEDs at the front and 12 reds at the rear are the base, one red at the rear and one yellow at the front are standlight. Both lights are built in old automobile foglights. The headlight has a window cut out in its red plexiglass, covered with transparent foil. The taillight combines dispersion through the glass with several diodes shining directly through holes.
Main switch S1 has three positions: off, lights (bottom branch on the plan) and USB (top branch).
Voltage for the lights is smoothed and accumulated by a set of capacitors C3 (three of them being in the taillight, the rest in the front) and limited by Zener diode D4 (theoretic maximum current for it is 41 mA, so no worries). White LED1 are connected in series of two, red LED2 in series of three, which means a threshold voltage around 6 V. Walking speed is enough to light it up. Standlight LED2 and LED3 drain the capacitors in about a minute after a stop.
This time I used a more sophisticated circuit for the USB charger: LM2576 switching regulator. Simple three-terminal linear stabilizers I've used before work as throttles: input and output currents are equal, just the voltage drops and the extra power is converted to heat. As if you pedaled as hard as you can all the time and regulated speed by continuous braking. A switching regulator, on the other hand, accumulates current in an inductor that works sort of as a flywheel. It either feeds it with full current, or cuts it off - as if you regulated speed by switching pedaling hard with coasting. The efficiency should be much higher.
How does the circuit work? Input voltage is limited to 36.4 V by Zener diodes D2; their power reserve is large enough to allow riding with unloaded charger on. C1 capacitor smooths out voltage ripples from the rectifier. The regulator measures output voltage by pin 4 and feeds the L1 coil by pin 2. When the coil "freewheels", the output current loop is closed via D3 diode. Switching frequency is 52 kHz, so this diode must open really fast to prevent dangerous voltage spikes from occurring - standard rectifier diodes are not enough, this one must be Schottky. Finally C2 capacitor (with low equivalent series resistance) smooths out the output voltage and we're done. The thickly drawn conductors should be built as short and thick as possible to avoid disturbing the regulation process by parasitic resistances, inductances and capacities.
The circuits are distributed over three etched circuit boards. Two "motherboards" are neatly bolted over the original light bulb sockets. Third one carrying LED1 is hot-glued to the remains of the red plexiglass cover. A note for the future: make the board bigger and mount it by the two screws holding the plexiglass.
Most of the interesting components are in the front light shell, the taillight contains just three C3 capacitors, LEDs and their resistors. The lights are interconnected by "faston" spade connectors. If I ever do this again, I'd put a protective diode between the positive and negative terminal that would short out the circuit if polarity is reversed, saving rear LEDs from destruction.
Front motherboard:
The most difficult part was mounting the big and heavy coil. I solved it by an aluminium sheet that also works as a heat sink for the regulator. To avoid scratching the coil and shorting it out, I added rubber pads made from an old inner tube.
The headlight before final assembly:
The tiny black brick bolted to the aluminium heat sink is the switching regulator, in case you're looking for it. All cables are doubled for safety purposes. Sides of the LED cavity are covered with silver paint for plastic models. The four screws in the corners support the hot glue blob holding the LED PCB.
Finished headlight:
The foil is glued by polyurethane glue and sealed with hot glue. All holes for USB and cables are located on the bottom side, so rain shouldn't be a problem (tested under a shower). The mount is made of steel sheet. The lonely screw in the opposite corner is one of the two holding the inductor in place.
Taillight:
Eight uniformly spaced LEDs illuminate the whole translucent backplate from inside, three "long range" diodes shine directly through holes (the middle one is standlight and the leads of the other two diodes don't touch it). Last two are partially embedded into the red plastic and point sideways. The black plastic shell is partially covered by aluminium sticky tape to improve its reflectivity.
Switching regulator circuit is more expensive and more difficult to design and make than a simpler circuit with a linear stabilizer. How do they compare?
I measured under pretty unfair conditions with the equipment that was readily available: 7805 linear stabilizer powered by a dynohub in a 28" wheel versus LM2576 switching regulator powered by a cheap bottle dynamo. I tested a range of speeds and loads; the most interesting output of the whole experiment are the following two plots:
Five volt stabilization with varying load is far from perfect, the voltage drops with increasing output current regardless of speed. In other words: you don't get above these lines even if you break your national speed limit during a downhill ride. In the usual working range, output of the two regulators is more or less the same.
This plot maps the bottom limit of stable regulation: if you slow down to the left of the lines, the voltage drops and the circuit doesn't work as a charger anymore. An area of stable voltage is to the right of the lines (disregard currents over 400 mA, the voltage may be stable there, but too small - see previous plot). As you can see, the switching regulator is more efficient and needs smaller input power (and speed) for the same output. But the difference is not very big.
Another conclusion is that the USB standard of 100 mA can be reliably generated from some 8 km/h upwards, regardless of the regulator type. The 500 mA standard can't be reached with this equipment at all.
I'm mostly satisfied with the result. The switching regulator is not such a wonder I had hoped for, but it works well. And the light output is excellent. Next time I'm going to do something similar.