Electronic – AC overvoltage protection for bike dynamo (alternator)

Your circuits (whether they work as you intend or not) seem to be based on the idea that when an over-voltage condition occurs, you will isolate the load with a MOSFET.

I think this is not a great approach. Consider, why did over-voltage occur anyway? You could say that it's because the impedance of the load is too high. Were the load impedance lower, then more current would flow, providing a way to sink whatever excess energy there is without the voltage becoming too high.

So, if you try to isolate the load with a MOSFET, that just effectively increases the load impedance. For a great many things that might have led to your over-voltage fault in the first place, they will simply respond by increasing voltage even more. This is problematic, because MOSFETs have relatively low maximum voltages. When a source-drain voltage is applied in excess of this, they go into breakdown and conduct (not unlike reverse breakdown of a diode). Consequently your load gets toasted anyway.

A different approach is a crowbar circuit. Instead of disconnecting the load, you short it out (usually with a TRIAC or SCR). If you put a short across the load, the voltage will be reduced (to 0V, for an ideal short). This provides very effective protection to the load from over-voltage.

Of course this also results in huge current drawn from the power supply. That blows the fuse, which isolates the power supply from the circuit, and does so much more effectively than a MOSFET can. Furthermore, the crowbar continues to conduct until there is no more energy which can result in a voltage across the load, which might take a little time, depending on just what caused the fault, and what reactive components are in the circuit.

Over volt protection is a common and accepted practice. There are many ways of doing this, but I won't mention them all. Firstly, you can place a big Zener diode across the output that is arranged to conduct heavily before the max voltage rating of the chips are exceeded. If the Zener eventually fails short circuit the chips are safe.

Secondly, you can place a big SCR across the output that is arranged to fire when an overvoltage occurs. This is a very old scheme that dates back to big linear regulators and mainframe computers.

Thirdly, what I do is place a cheap, low Rds(on) N-chan MOSFET in series with the output of the buck converter. I drive the gate of the MOSFET with DC that comes from the converter switching waveform. I use a simple diode pump to do this. If the main buck converter FET fails short circuit which is the most likely failure mode the diode pump stops and the power is cut off.

There are also converter topologies that don't give an overvoltage when the switch fails short circuit but they are not as simple and as efficient as the buck converter.