Is Reply # 1234 in a foreign language?Too clever for me all that lectrickery stuff :& it bites 
Oops, sorry about that. The cunningness here is that instead of the standard method of using
diodes to convert the alternating current from the dynamo to DC, this is using
transistors to effectively 'connect' the dynamo only when it is producing a higher voltage than the capacitor (which is acting as a small rechargeable battery) has stored. As the dynamo output drops, it is 'disconnected' to prevent the stored charge being wasted by flowing backwards through the dynamo. The transistors work in pairs to 'connect' the dynamo 'the other way round' on every half-cycle, so the AC is converted to DC in the process. The capacitor stores charge, smoothing out this pulsating voltage and providing a reserve to keep things going for a bit while the wheel is stopped (as in a standlight).
This has the advantage that the transistors, unlike diodes, do not 'drop' significant voltage. This means that the full output voltage of the dynamo goes into the capacitor, which means the regulator can continue to provide a valid 5V USB charging voltage at lower speeds (and I'm all about the low speeds). It also means that hardly any power is wasted as heat in the AC-DC conversion process.
For additional cunning, I'm not using
your standard regulator. Instead of wasting the difference between its input and output voltage as heat, this one
switches on and off rapidly, in a manner not entirely unlike the first half of the circuit. This is much more efficient (the
device claims 92-94% efficiency), and makes nearly all of the dynamo's output available for powering something useful (in this case, a USB phone charger).
Make sense?
Since capacitors and electronics, like the rest of us, don't like being exposed to overly high voltages, there are a pair of
zener diodes limiting the dynamo's output to 16V, in the same way they'd be used to prevent a halogen lamp from blowing at high speed. The higher this 'clamp' voltage, the more power can be extracted from the dynamo. Unfortunately, the limiting factor here is the regulator, which accepts a maximum of 18V on its input, so I can't actually increase it by that much.
I'll handwave over the bit where I made a rookie mistake in using an oscilloscope.
Next step is to work out the best way of politely charging potentially odd numbers of NiMHs (I have a head torch that takes 3 cells, but lights and GPS take 2) from an unreliable power source, possibly a bit of control logic for load shedding (keep the GPS supply going to the bitter end, but stop charging batteries if I'm puffing up a hill and there isn't much power available) and of course all that awkward physical construction stuff. Also investigate solar panels, as it would be trivial to add one as an additional power source, and that might be useful for camping trips where less time is spent in the saddle.
