Hi,
[See attachment]
Well we have to remember that if the voltage differential across the regulator is too high it wont put out the full current either. That's why the last test i'd like to see is where we include the regulator as well as the bulb, and gradually raise the input voltage to the regulator (LM317 or similar with higher current ability). As we start to raise the input, we should see the bulb get more and more current and start to heat up, and with a limited differential voltage we probably see it light up to full brightness as the input gets to around 14 to 15 volts, then as we go above that 16, 17, 18, i think that eventually we will see the bulb turn off because the regulator starts to limit current just based on the voltage differential alone. And once the current in the bulb decreases, the resistance again decreases, and the regulator limits even more and more and so the voltage differential continually increases and this 'locks' the regulator into a low value current limit that could be a lot lower than it's normal current limit. This would keep the bulb off indefinitely until the circuit power was removed and then replaced again, but of course it would turn off again.
A related test would be to simply see what current it puts out with a large voltage differential. Tune the output for say 3v and use a 3 ohm load perhaps, and see what input voltage causes the current to go down in the output load.
So the first current limit is the value of the rated current limit and that is due to a pure overcurrent load, but the second current limit is due to a high voltage differential and this current limit will not reset automatically but will latch because the bulb resistance goes down and that causes an even higher voltage differential.
Looking at the diagram, we can see that if the bulb started normally it would probably run ok if it only drew about 1.8 amps nominal, but if the bulb didnt start first then it depends on that curve whether or not the bulb starts. It does look like it would start but it might be too close to call. [If the bulb didnt get hot at all the regulator would see V2, but if it started ok the regulator would see V1, so during startup the regulator diff voltage follows the curve from V2 to V1 progressing to the left].
At some point maybe i'll see if i can include this behavior into my model for the regulator and test it with my model for the bulb.
This is what i believe could be happening, barring a simple wiring error. I dont think the spice models model that aspect of the device either so we'd have to use a real life device to test this theory.
[LATER]
Ok, i threw together a quick behavioral model of the secondary current limit effect, modeling the curve as an offset quadratic for any differential voltage above 11v because that's what the curve looks like in the data sheet. It's an approximation but i thought it might tell us something about the behavior even if it wasnt dead on accurate. The 'normal' current limit was set to 2.6 amps so it would be above the 11v secondary limit point which was set at 2.2 amps.
What i found was that with input voltages (to the regulator) less than some max level, the bulb voltage would rise slowly at first and then a bit faster until it went up really fast (very steep slope) and then it would reach max (12v for a 12v bulb).
Raising the input voltage the bulb would turn on slower and slower, but still reaching a point where the voltage rises very quickly until it reaches 12v, so the bulb still started ok.
But finally reaching some set level of the input voltage (in the case of this actual model set it was at 28v) the bulb current would rise slowly as usual but never reaching the point where the voltage started to increase more quickly. In other words, the voltage would taper off to some set level and never rise above that, so the bulb never started completely.
Then testing it with no time constants (thermal time constants all set to 0 so it would take no time to start in a real setup) the bulb still would not start at 28v.
The reason for any lack of start condition was always due to the secondary current limit internal to the LM317 type regulator, which is based on the voltage differential across the regulator.
Note that other regulators might give different results though if they have a different curve than this one does.
The difference between 24v and 28v is significant, but even with that spread it's hard to say if the regulator was a little different or the bulb a little different the limiting action might occur at 24v instead of 28v. We'd have to test a real regulator/bulb combo.
What else this says is that the regulator doesnt need time to heat up in order to see this limiting effect. It's not the thermal limiting it's the voltage differential limiting. That's not to say that it doesnt eventually heat up anyway though and start to limit because of that too.