I plan to build a LED driver that can pulse a high-power LED (3 W) down to the 100 s of nanoseconds pulse regime, yet also allows longer pulses (up to 100 ms) should be possible. The LED should be driven off a TTL/GPIO signal.

I’m admittedly a bloody beginner in electronics but did some reading upfront. I came to the following conclusions so far, which also might be wrong:

  • Buck/Boost driver are too slow,
  • Timer or avalanche circuits have fixed pulse width,
  • Single MOSFET driver can introduce oscillations,
  • My best bet is a push-pull MOSFET driver,
  • Shunt drivers might also be an option.

Sadly, I haven’t found a comprehensive tutorial on designing such a driver specifically for driving high-power LEDs. If there is one I have overlooked I would be happy if someone could push me in the right direction. If there really is none, I have the following questions:

  1. Which transistor pair would be good to drive a 3 W LED in push-pull?
  2. How do I calculate the resistors specifically for the LED (3.5 V, 700 mA)?
  3. What type of power supply can I use to drive the circuit? Are voltage regulators interfering in such circuits?

Independently from this: What do I need to do to step such a circuit up to drive a 30 W LED?

As mentioned above I would be happy to read a existing tutorial about this if this is a repost!

Thank you!

  • 1
    \$\begingroup\$ It's not possible to pulse <1us with power LED's due to high capacitance. not even close. But you modulate a Laser after it activates. Consider the ESR*Cjcn =T value. 1A white LEDs need a good heatsink but Vf ~ 2.8V+If*0.5Ohm +/-50% \$\endgroup\$ Dec 5, 2016 at 23:48
  • \$\begingroup\$ @TonyStewart.EEsince'75 - White LEDs are not suitable for microsecond pulsing due to phosphor persistance. \$\endgroup\$ Dec 5, 2016 at 23:57
  • \$\begingroup\$ that's right about 8ms decay time \$\endgroup\$ Dec 6, 2016 at 0:03
  • \$\begingroup\$ Start with a buck circuit with hysteretic control, do not use output capacitor, use a shorting/shunting switch across the LEDs. That can comfortably produce a regulated current pulse of a few hundred ns. In particular, look at the datasheet of LM3409, there are examples and waveforms showing the fast current switching. \$\endgroup\$
    – rioraxe
    Dec 7, 2016 at 6:20
  • 2
    \$\begingroup\$ What is your app? High-speed photography? \$\endgroup\$ Feb 18, 2017 at 9:52

1 Answer 1


First read this: http://www.osram-os.com/Graphics/XPic5/00135349_0.pdf/High-Speed%20Switching%20of%20IR-LEDs%20(Part%20I).pdf

I'm not sure of your application, but 3 W LEDs typically have dreadful turn on characteristics, you'd be lucky to get 800 nS tr for many of them. But it depends what you want to do; if your doing IR based TOF, then you need to move to multiple smaller LEDs but if it's just for a low bandwidth communication channel, or photo flash it may work out ok.

To get the best possible switching speed for whatever LED you select you need to bias the LED into it's knee and you can do that like this (though this method consumes power all the time):


simulate this circuit – Schematic created using CircuitLab

I've seen push-pull drivers, but not that work well; they clamp the LED to 0 V which means you have to charge the junction right from zero limiting turn on times drastically as the chip size goes up.

  • \$\begingroup\$ By the way, this will not scale to 30 W and provide 100's nS pulses. 30 W LED sources are invariably multi-chip and so the drive voltage will increase with LEDs in series. This makes achieving fast rise times more difficult. \$\endgroup\$ Dec 6, 2016 at 0:19
  • \$\begingroup\$ Thank you for this answer! They circuit seems to be pretty straightforward, yet what is the reasoning for placing R2 at the place it is? What do you think of an IRL540 as a mosfet in this scheme? \$\endgroup\$
    – JJCale
    Dec 7, 2016 at 21:20
  • \$\begingroup\$ The highpower LED I was talking about are the Luminus CBT-90 color LEDs which apparently accept 13.5A at 3.4V on a monolithic chip. Could I drive this LED in the same manner you sketched up? I don't care about efficiency. \$\endgroup\$
    – JJCale
    Dec 7, 2016 at 21:27
  • \$\begingroup\$ @JJCale. You certainly could scale it up to larger currents with the right components. Though I'd suggest that instead of a resistive high side drive you might want to build a constant current high side driver. This get's your driver power dissipation into a heatsink and provides better current control for variations in device Vf. I notice from the datasheet the Vf varies from the knee at 3.2 V to around 4.2 V @ 13 A. So you might be able to use two series connected Vishay FES16HT-E3 diodes (high voltage fast recovery) to get close to the 3 V knee when turned off. \$\endgroup\$ Dec 7, 2016 at 22:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.