Electronic – Advantage of using an op-amp connected emitter follower for LED driving

This circuit will work, but not be efficient, power-wise. If the output is high the LED is on and you'll have 20 mA current through LED and resistor.

But with the output low, the LED is short-circuited and therefore off, but the resistor current will be even higher: 50 mA!, which is not only bad for your wallet and the environment, but also for the output pin; most I/Os will only allow 25 mA.

Place the LED in series with the resistor. Then a high output won't light the LED because the output FET is switched off. A low output will draw the 20 mA through LED, resistor and open-drain output.

Note that this will invert the logic: in your first schematic the LED will be off when the output is active, in the second circuit it will be on with active output.

edit
If the output can't sink enough current you'll need an external transistor to increase that.

If the output is active (low) there will flow a 1.5 mA from +3.3 V through Q2's emitter-base junction, R2 and Q1. That 1.5 mA will allow Q2 to source more than 100 mA for a typical general-purpose transistor. You will only get about 22 mA, though, because R1 won't allow more.

If Q1 is off there won't be any base current through Q1, so the LED will remain off. Q1 will have a small leakage current, and to avoid that this would get amplified by Q2 I added R3. As long as the leakage current is less than 0.7 V / 15 kΩ = 50 µA all the leakage current will flow through R3, so that will ensure Q2 will be completely off.

Here's an overview of the design process to get you started. I'll let you work out the exact calculations.

I would replace \$R_{\text{load}}\$ with an independent current source \$I_{\text{load}}\$ for your simulation (you can use your CircuitLab schematic for simulation once you add resistor values). Set \$I_{\text{load}} = 25\$mA since that is your worst case.

Pick a relatively large emitter resistor \$R_3\$. This simply provides a load to the transistor if the actual load isn't connected (e.g. \$I_{\text{load}} = 0\$). For example, use \$R_3 = 10\$k\$\Omega\$. If \$V_{\text{out}} = 5\$V then the current through \$R_3\$ is \$0.5\$mA and \$I_{E} \approx 25.5\$mA in the worst case (\$I_{\text{load}} = 25\$mA).

Next you need to determine the worst case (highest) \$I_B\$. Use the lowest \$\beta\$ in the transistor's datasheet (worst case) and then calculate

$$I_B = \frac{I_E}{\beta + 1}$$

Now in order to make the resistor voltage divider "stiff" you need to make sure that the unloaded bias current through the resistors (call it \$I_{\text{div}}\$) is at least 10 times the load current (in this case \$I_B\$ is the load for the voltage divider). Otherwise the load current draws too much current away from \$R_{2}\$, which causes the voltage at the output of the voltage divider decrease too much. This puts a constraint on the maximum value of \$R_1 + R_2\$ since

$$I_{\text{div}} = \frac{15}{R_1 + R_2} > 10I_B$$

This equation plus the voltage divider equation

$$\frac{R_2}{R_1+R_2}15 = 5.6$$

gives you two equations and two unknowns.