How did you choose 12Ω for the load? That would be close to half an amp at 5V. Perhaps the real load is less.

Also, how do you know what a "reasonable voltage" is? The DC-DC converter might be fine with a larger ripple than you think.

6V, 12 Ohm, 3W. That's what the dynamo is specified for.
As said, I'm looking for the highest possible efficiency, since every few extra watts have to be pedaled for. I agree, you don't really feel a few watts, but believe me, after 10, 20 or 40 hours you will!
At the same time it should be simple and tiny...

 

6V, 12 Ohm, 3W. That's what the dynamo is specified for.

Well OK but isn't most of that capacity going to your lights? I thought only a fraction was going to the DC charging circuitry. What current do you want from your 5V USB port?

 

You might consider not filtering the AC and see how your buck-boost circuit handles the rectified sine-wave directly.
Your rectifier with MOSFETs shown as the second circuit in Post #2 would work for that.

 

A caution about voltage multipliers: They mostly work by storing power on one capacitor during a positive half cycle and on a second capacitor on the negative half cycle. And the caps are in series. So with a low peak rate the capacitor storage gets rather demanding.

True, but you can have too much capacitance too. Increasing the capacitance increases the magnitude and shortens the length of the current pulse to charge them. The losses are proportional to the peak current.
The old-fashioned "bottle" "dynamos" (they were alterators) were made with lots of series inductance so that the speed increased so did the voltage, but so did the output impedance, sort-of crudely regulating the output.

 

6V, 12 Ohm, 3W. That's what the dynamo is specified for.

Well OK but isn't most of that capacity going to your lights? I thought only a fraction was going to the DC charging circuitry. What current do you want from your 5V USB port?

when using the lights, at night, it's 100% for them and no recharge. During the day it's the opposite.

 

when using the lights, at night, it's 100% for them and no recharge. During the day it's the opposite.

Ah, OK, so you want to design for the full 3W (ignoring losses) at the 5V port, so roughly 1A to be safe. Got it.

I maintain that a lot depends on the DC-DC converter you choose, and what it can handle on the input while maintaining 5V on the output. You can't really design the input power section until you know the specs that device requires.

 

Warning: Not my field of expertise!
- in a distant past a ( germanium diode?) rectifier for an AC milliVolt meter input probe was pre-biased to just under conduction-
Idea: fit a small ( nF?) capacitor between (Logic 2V Vgs) mosfets S-G, pre-charge it to 1.8V or Vgs minus 0.1V ) with a schottky diode from the generator and limit Vgs to a safe value above conduction, find a way to discharge cap to just below conduction at AC zerocrossing, so opposite input voltages are rectified - - who will pick up from there ? simple should do it. Have fun !