tlf9999 said:
I don't know why Q2 will be killed - I think Q1 has a better chance of getting killed.
The "flaw" in the design is that the voltage drop over Q1 is always Vbe of Q2, and if there is a lot of current going through Q1, it will get hold quickly.
Also, the design would be a lot more interesting if it is done on N-channels.
Not true on either account.
It's a Pch MOSFET shich means that the resistor R3 (or Rg, difficult to read) will keep the Vgs equal to Vin unless Q2 prevents it. For standard VMOS this implies an absolute maximum Vin of 20V. A zener diode between G and S to limit the Vgs will alow higher input voltages.
Vds will be nigligible as long as Vin is over some 6-7V, because the MOSFET will be working in the Rdson mode, as a very low value resistor. Q2 senses the voltage drop on this resistor and if it approaches 0.7V, turns off the MOSFET. As Vds will then increase (MOSFET goes open), Q2 will continue to keep the MOSFET off unless the current drops to a value where the voltage drop on R1 will become lower than 0.7V.
And here is why Q2 will die, almost instantly, unless it gets a base resistor:
Assume voltage source at input and short at output, for instance an output capacitor - in order for the MOSFET to start conducting, the input voltage should be at least 3V or so, unless a very low threshold MOSFET is used, in which case it can be about 1V. This implies 1-3V across BE of Q2, which pretty much means instant death. R1 must be chosen such that R1 never develops more than 0.7V across it until the input voltage reaches 3-4V, which is a problem without a proper solution - for a short circuit at the output, the input supply would already have to be current limited for this to work, which begs the question why this circuit in the first place.
Even if Q2 survives power-up, when the voltage across the MOSFET reaches 0.7V, R1 would then have to somehow keep it limited to 0.7V for Q2 to survive. Once 0.7V is reached across BE of Q2, the MOSFET is switched off, which removes one path for the output current, and leaves only R1, which means that the voltage across BE of Q2 would start suddenly increasing further (it's a foldback effect), for a short at the output, up to the full Vin can get across BE of Q2. Result: dead Q2.
As for Q1 heating up, since the voltage across it is limited to 0.7V or so, it comes down to chosing the appropriate MOSFET, Pdiss = 0.49/Rdson. This is actually quite easily manageable.
... but we still use mosfet's backwards at work for reverse polarity protection. 
