A number of companies make resistors with a very high temperature coefficient; when too much current is passed through them, they heat up to the point that their resistance increases substantially. When no more current is flowing through them, they cool down and the resistance returns to normal.
types of FETs have two additional useful properties: (1) they have very low resistance until the current through them reaches a "pinch-off" value, whereupon they will current-limit; (2) the pinch-off current drops with temperature.
If one could construct a module with a JFET or similar device combined with a thermally-sensitive resistor and possibly a couple other components, it should be possible to arrange it so that it would pass currents below the threshhold with minimal voltage drop (and thus minimal heat dissipation). When an over-current condition occurred, the device would heat up and reduce its pinch-off current to the point that the device could sustain a voltage drop without overheating to the point of destruction. As with a polyfuse, removing current would allow the device to cool down and return to normal.
This device would have two main advantages over existing "self-resetting fuses":
-1- Existing SRF's tend to have very "mushy" performance curves. The amount of current required to make the device trip out in one second in a cold environment is much greater than the amount of current the device can pass continuously in a hot environment without eventually heating up enough to trip. Because the FET-o-matic fuse would have minimal resistance as long as current remained below the pinch-off point, there would be less difference between the "quick trip" and "never trip" current levels.
-2- Existing SRF's provide limitted short-circuit protection: a 250mA SRF may conduct many amps for a few ms before tripping. The FET-o-matic unit, by contrast, would never pass more than a certain level of current. The amount of power dissipated in the FET-o-matic fuse may be very great under short-circuit conditions (e.g. a fuse that limits current to 4 amps on a 40-volt circuit would generate 160 watts of heat) but if it was properly designed it would only generate this much heat until the substrate reached a temperature sufficient to force a cut-out.