The attached schematic illustrates the principle: use a 555 running at fixed frequency with variable duty cycle to drive a N-FET tied to ground. Connect the FET to the LED using a series resistor. As you vary the PW, the LED brightness varies. FET stays cool because it never operates in linear region (power resistor burns the heat)

The circuit shown could be simplified since you don't need to vary the frequency, just the PW. Connect the neg LED pin to the + output shown in the schem,

You would NOT need a current source at the top of the LED, just tie it to V+. The varying PW varies the average current through the LED which sets the brightness.

 

See if I get this right. U1 is configured astable and _sets_ the flipflop in U2, then U2 switch itself off after a while ?
The output circuit is interesting. Are those parts around the mosfet snubber circuit like they talk about in the other thread for reducing dV/dt ?

There is a 555 based PWM circuit in another thread (and here). It reduces to only one 555 chip. I've made an atempt to use it together with a similar setup but with lower frequency for blinking mode. What do you think ?


please ignore the values and switchfrequency for now and the output stage.

I don't like the resistor-led solution here. The resistor value (<1 Ohm) would now be getting closer to the sum of resistances in the loop (battery, connections, wires). The other thing is led current will vary much from full to low battery. I'm planning to replace that resistor with a multiple of AMC7135. I will get 200mV dropout, but with the XM-L having Vf close to the cut-off voltage of the L-ion, it should stay in regulation a long time.

 

One 555 is free running astable to trigger the second which has variable ON time (pulse width).

The adjustable R-C is to make the rise and fall times more symmetric (not needed for you).

R9 and C6 are snubbers to reduce ringing on the FET.

 

Is there any reason you can't just use your current source, and add 2 resistors?

If needed you could even tweak the 2 resistor values to suit the specific LED Vf values.

 

What you are doing is putting the leds out of balance without any regulation. Lets say you set the balance between the LEDs to 25/75 at startup, then LED1 with higher current heats up more than the other. Depending on how good the thermal path between LED 1 and 2 is, LED1 will now conduct even more current when its hot, and as a result LED 2 will conduct less , increasing the difference in temp between the two even more. You get a thermal positive feedback. I can demonstrate this just for experimental when I get the components and make a short video.

2nd. You won't get any realtime adjustments.

When I get my stuff and will build, should I continue in this thread or open a new one?..

 

The 2 resistors will compensate enough of the thermal characteristics of the LEDs, provided they drop more than 0.4v or so. It is a proven and well understood system. Google "emitter balancing resistors" if you need to learn more.

So if you already have your 2.8A constant current source, and 2 LEDs, then just add 2 resistors. Problem solved.

An alternative would be to use 2 cc regulators, like 2 LM317 regulators, one drives each LED. This gives an added benefit of redundancy on your vehicle light, so if a regulator fails completely you still have one LED working perfectly as opposed to the disaster you get if you use a single regulator and it fails.