User power supply current limit to avoid current limiting resistors for LED

The advantage of an incandescent light bulb over a resistor is that its cold resistance is about 10 times less than when lit up to normal brightness. Therefore you can choose a wattage that limits short circuit current to a safe value, without excessive voltage drop at lower currents. Also, since it takes some time for the filament to heat up and increase resistance, short duration current surges cause much less voltage drop.

You should choose a lamp with voltage rating close to the power supply voltage (a little lower won't hurt) and wattage that gives an acceptable short circuit current. Power = Volts x Amps, so if you wanted to limit a 12V supply to 0.5A then you would need a 12V 6W lamp.

The maximum normal operating current will then be determined by how much voltage sag is acceptable. This is tricky to calculate because the lamp's change in resistance with current is very non-linear. A 12V 15W festoon lamp that I tested had a resistance of 0.9 Ohms when cold, and 9.8 Ohms at full brightness. At 0.3A (24% of the bulb's nominal current draw) the output voltage dropped from 12.0V to 11.5V. Here is a graph showing output voltage vs current for my lamp compared to using a 10 Ohm resistor.

The 'soft' current limiting that a lamp provides is good for protecting against continuous short circuits without affecting normal operation, but won't protect sensitive electronic components which cannot withstand short current surges or slightly higher than normal operating current.

If you need a fast acting current limit with sharper cutoff then you could use a bipolar transistor (PNP for positive voltage, NPN for negative) with its Emitter connected to the power supply, Collector to the load and a high value resistor going from Base to Ground. Maximum collector current will then be determined by the Base current multiplied by the transistor's current gain (which varies from unit to unit, so you may have to adjust the resistor value to get a precise current limit).

You are more or less right, but this is not a good solution. Your voltage at the collector of q2 will never get much above 1.4v. And your 190 ohm R1 will dissipate 30x0.15 = 4.5W

Because you have feedback you can use a fet to pass the current and make R1 much bigger. You will probably have to use heat sinks on Rsense (1W+), and your pass transistor q1 will have to take the full whack of power. That is near enough 45W. Also, your limit will vary by at least 300mA across the temperature range, maybe 1.3A minimum to 1.6A maximum.

There are many more efficient and more accurate solutions. What other design criteria do you have? Eg Cost, Accuracy, board space?

Followup to comments: Below, Enrico asked to see some alternatives. One possibility is to use Lmp8646 which is a current sense amplifier from TI. You would need Rsense of about 0.1R. Set the gain accordingly then feed the output into a comparator. Use the comparator to turn off the supply (does your regulator have an enable input?). For this circuit you would need a fixed supply less to power the ic, comparator, and reference. The resistor would dissipate just under 0.25W. I would suggest putting a lot of hysteresis on the comparator and also a capacitor on the output to stop the thing switching on and off very fast during fault condition. You could instead latch the supply off until a reset button is pressed.

This IC is also suitable for more sophisticated current control, such as variable current limiting and folding back the supply voltage by connecting the output into the regulator feedback. It is just the first one I came across - other brands / circuits are available.