A DC generator is used to charge a supercap bank, rated 10 farads @ 20V. The bridge rectifier is only there so that regardless of clockwise or counter-clockwise rotation there's always +/- as depicted. I wanted to establish a shunt controller to limit the charge on the cap bank to 18V.

When Vin reaches 18V, Q1 turns on due to Dz1 having a zener voltage of 18V. At first I tuned the circuit without M1 connected, and established 10V at the wiper of pot R2 when Q1 is on. 10V ensures M1 turns fully on ... or so I thought.

I connected M1's gate. As soon as Vin reaches 18V, M1 indeed turns on. However the unexpected side effect of this is that now due to the extremely low impedance in M1's path (1 ohm + 0.0092 ohm Rds), Q1 cannot sink current enough to keep Vgs > ~4V, or barely above threshold. M1 never turns fully on. The result is all the "extra" current being pushed out by the DC generator is dissipated in M1, instead of my shunt load.

Is there a different approach I should use here to limit the charge on this cap bank? The DC generator can basically be thought of as wind driven.

I tried to adapt a solar-panel diversion controller to this supercap bank, but the idea doesn't quite work with supercaps due to their ability to source current very fast. I tried to use

http://www.redrok.com/electron.htm ( look for "Diversion 1" ) as my template. The above schematic was taken from this URL.

Any suggestions?

 

Completely wrong idea.

If this was the right circuit, you would have to provide some leakage current to the zener diode so it doesn't come up to voltage slowly. 5ma from emitter to base of Q1 should do it.

.55 volts/.005A = 110 ohms. 100 or 120 will do just fine.
Then short out the 2.2k resistor.

This gives Q1 a solid switching point. The theory is that when Vcap >18.55V Q1 will switch on hard and R2 will set the applied gate voltage to 10 volts.

The problem is that M1 will then regulate the voltage to 18.55 volts, reduce the current through Q1, and become the variable resistor that suffers the heat load.
If you want the first circuit to work, you will have to add components to let it overshoot, turn off, and start back up. It will become a pulse frequency modulated load dump. Either that, or believe M1 will suffer the heat load and provide a proper heat sink for it.

I don't think I can design a pulse frequency modulated load. Maybe somebody else can?

I didn't even think about the second schematic.