Why this happen

This circuit works with all voltages. What this circuit doesn't do is guarantee that it won't blow up your LED. See, what you have is a current MIRROR and not a constant current source. A current mirror merely "copies" the current that flows through the biasing transistor; the current flowing through the biasing transistor is a function of the input voltage as a resistor is used to set the biasing current. So I(bias) = V(in)/R(bias). At 10V, I(bias)=25mA, which doesn't blow up the LED. For higher voltages, the bias current is higher. Your circuit is equivalent to a resistor in series with the LED and the transistors are redundant.

As yoy said,this circuit should work at all voltages, but in my test, right transistor current does a sharp increase when source voltage slightly goes up from 10V. It seems that in my tests, for a source voltage > than 10V, the circuit doesn't works like a current mirror anymore and suddendly starts flowing much more current than it should throught right transistor.

I forgot to mention that i tried different versions of this circuit with the same results, with different transistors ( 2907 pnp, bc336, 2n2222a, pn2222abu) and replacing the led with a resistor and remounting all components in the breadboard

@cmrincon, if you touch right transistor by hand while increasing voltage you'll sence it gets extremely hot when voltage exceeds certain value. These two transistors don't work in the same temperature mode. Extreme scenario cause thermal breakdown.

@schernya yeah, as you pointed, hfe of the transistor depens on the temperature, the more temperature, more hfe. As transistors aren't exactly the same, hfe is not the same and therefore i can't guarantee that the right transistor won't cause thermal breakdown. This is not my case. Transistors had not been working for more than a minute, even i took the precaution of increasing the "adjust resistance" to 10k which gives me a dissipated power around a few milliwatts. That's too low to make that effect that fast. I thought that probably transistors were faulty, but i get the same result with different transistors, even i changed to original pn2222abu transistors, the only difference i noticed was that this effect started at a voltage of 17 instead of 10

@cmrincon, thermal breakdown is your case exactly. Check collector current with a meter and you'll see that it's way to far from expectations. Vbe of the right transistor is fixed, but emitter resistance drops with cathode current increase. And it will increase more than one may expect because relatevely small load sits in the collector of the right tranisitor. That's what happens when you increase collector voltage.

@schernya doesn't makes sense that what you say, is a slow process which makes current flowing throught right transistor to increase as temperature also increase, instead of a sudden increase of the current? Because i notice this slow effect in the transistor, as temperature increase.But is really slow

@cmrincon it is slow effect if you keep collector voltage constant. From your description I understood it's not. If it doesn't make sense, take a meter and sure enough you'll see it yourself.

Try this one. It rectifies sudden emitter resistance drop and thus thermal breakdown. http://everycircuit.com/circuit/6452376972820480

I think that this may work too https://everycircuit.com/circuit/5854917460492288 thanks you

The voltage across this transistor is Aprox. 10V and the current 22 mA, then the disipated power heats the transistor. Eventually change the beta factor and heats even more, finally burns the transistor