I need a solid-state switch to pass a charging current or block it entirely:

The current source is a permanent magnet DC generator. It is charging up a 35 farad, 18V supercap array. When completely discharged & the switch closed, the cap looks like a dead short to the generator. Consequently the cap + Rload draws at max about 2.5A. As the cap bank charges up, the charging current drops as expected. When the charge on the cap bank reaches 18V, I want to completely shut off the charging current by opening the solid-state switch; this is to prevent over charging the caps. For arguments sake, let's say the DC generator is driven by a person on a bike, and they keep pedaling long after the cap bank reaches the limit of 18V, for they are unaware of anything to do with the state of the cap.

I thought at first to use an n- or p-channel power MOSFET to realize the switch, but because I have no stable voltage to work with ( other than the voltage developed on the cap bank over time ) I cannot turn on the MOSFET: n-channel Von is ~5-10V but at t=0 (caps fully discharged) I have zero volts to work with so n-channel cannot be on. For p-channel, I could start with the gate grounded, but I'm not sure how I would develop maximum voltage at the gate to turn it off once the charge on the cap reaches 18V.

Does any one have a suggestion as to how to realize the switch above, ie, have it be automatically closed at t=0 when charging starts and then open once the charge reaches 18V?

 

...When completely discharged & the switch closed, the cap looks like a dead short to the generator. Consequently the cap + Rload draws at max about 2.5A.

Not true. A dead short draws about 1.21 gigawatts.

It's a short.... there is nothing to limit current.

Next, 18 volts on a 35 F 18 volt cap is gonna make one BIG explosion when the cap shorts and dies due to excessive voltage. You need more margin there.

 

The current is self-limited due to the construction and the design of the DC generator. It'll put out at max 2.5A at 1000RPM and at t=0 when the cap bank is fully discharged. That's a little bit over-rated but I'll deal with that later. The cap bank is actually rated to 25V; sorry I didn't make that clear earlier. So I want to clamp it at 18V and then disconnect the DC generator.

How can I do this with no stable voltage reference? Moreover, how can I do this without a relay, relying on solid-state electronics only?

 

The open-circuit voltage of the DC generator is about 120V; so when I flip a mechanical switch open, the voltage spikes at the generator output. I thought of using a power PFET to pass the current to the cap/load, but because the cap bank is max 18V I have no way to generate the total open-circuit voltage of 120V to turn off the PFET. Still stuck ...

 

The current is self-limited due to the construction and the design of the DC generator. It'll put out at max 2.5A at 1000RPM and at t=0 when the cap bank is fully discharged.

This may be true but don't forget that transcients can occur [and they usually do] and when they do, current will surpass the 2.5A limit as Ernium points out...unless the gen is internally clamped which is somehow wasteful IMO.
Transistor-based voltage regulators will do. Theraja has something about this...
Consider using a power NPN transistor as in this link:
http://www.circuitstoday.com/wp-con...mitter-Follower-Voltage-Regulator-300x169.jpg .
This series network works best because the overall power dissipation is much lower than those of shunt networks.
Do the math and remember a heatsink if necessary. Put your Nmos switch before this voltage regulator circuit.
Cheers

 

The 2.5A current limit isn't the issue in this thread; I've found current limiting strategies similar to what you posted, m1ch431 albeit using 2 NPNs or PNP's cross-coupled and they work OK.

The issue is: how does one realize the solid state switch for disconnecting the DC generator from the load? It has to block at least 120V, which is the open-circuit voltage of the DC generator at a specific RPM. I can't find any SSR that are normally closed that will block 120VDC. Therefore I figure I have to build one myself.

 

Here's an idea of what I'm trying to do:

U1 is a comparator that is tuned via R2 & feedback R7 & C2 to flip at Vth=22V and Vtl=20V. U2 serves only to invert this signal. The inverted signal is applied to the base of Q3, turning it on. R11 and R12 serve to limit the applied Vgs to Q2 to be 10V max when V on C1 is at the max the circuit is tuned to (22V).

I haven't built Q1 & Q2 into the circuit yet ... but I am wondering, am I on the right track here? I have to allow current to flow into C1 until its voltage reaches 22V, at which point I want Q2 to pull Q1's gate to ground, turning it off. Q1 should be able to stand the 120V when it is off because Vds(max) = 200V for this part. However, if Q1 is off and the gate is at or near 0V, now I'll have Vgs=-22V and that will probably fry the part as Vgs is rated to +/- 20V.

I am not sure how to get around this problem.