I'm trying to make a fast LED strobe for visual periodic measurements of mechanisms, rotating objects to be specific. I "designed" an opto-coupler/MOSFET circuit and a transistor/MOSFET circuit. I plan to have an Arduino as the precision strobe signal source so i want to limit the current drawn from arduino digital pins to +-10 mA. After simulating both the circuits in LTSpice, i saw that i have to trade between base current vs. response times or peak load current. Maybe there is a proper way to design such circuits but I'm no expert. My goal is to strobe the LED with a minimum of 100 us pulses in 1000 us periods. I aim to drive a high power COB led that draw about 1 A in 36 V (40 W rated) so I could make an easily visible strobe.

Then came to my mind is that, why would I not be able to use an audio amplifier? The switching frequency i need is around 10 kHz (1 s/100 us), and audio amplifier input impedances are fairly high so Arduino digital output would have suffice.

Is it possible to strobe drive an LED with audio amplifier, with Arduino made pulse signal in the line input? If yes, how would I know the voltage gain of an audio amplifier which is rated by watts at certain ohm loads, so i could design a proper output voltage to drive COB LED?

Edit: schematics and LED current response added, the LED is modeled as a 40 ohm resistor (R2): transistor-mosfet driver R2 load current

  • 1
    \$\begingroup\$ "LED is modeled as a 40 ohm resistor" Please don't. The Vf will have a significant role here. \$\endgroup\$
    – winny
    Sep 6 '19 at 8:21
  • \$\begingroup\$ okay i realize it's not pretty accurate but i dont know any way to incorporate the i-v characteristics of the specific led i plan to use to ltspice. the rise and fall curves would have been steeper in an actual led, and this is the only difference that would take place within my limited knowledge \$\endgroup\$ Sep 6 '19 at 8:30
  • \$\begingroup\$ No need. I usually just cheat with ideal diodes in LTspice, or a combination of silicon and Schottky to match the Vf I'm after. My point is that any voltage below the Vf will get you zero current and thus zero light. \$\endgroup\$
    – winny
    Sep 6 '19 at 8:32
  • \$\begingroup\$ Okay, i will try to tinker with ideal diodes. Any voltage below Vf giving me zero current would actually cause the pulse to be shorter, i'm aware of it and it would benefit me in the application. But also in reality, the COB LED i plan to use starts to draw current at 33.3 V and goes up to 38 V at absolute maximum ratings. I wouldn't be completely satisfied with the results unless i model the i-v characteristics as given in the data sheet so i thought resistor would be a crude enough assumption. \$\endgroup\$ Sep 6 '19 at 8:37
  • \$\begingroup\$ Good that you are aware of it. Put a resistor in series to stay within maximum specified peak current at all output levels. \$\endgroup\$
    – winny
    Sep 6 '19 at 11:18

I think you will find that an audio amplifier is a poor choice as an LED driver.

It is AC coupled so the output swings both positively and negatively with respect to ground. Therefore your LED will have to be protected against the reverse half cycles if it is to have a long life.

An audio amplifier output is a low impedance voltage source, LEDs require a constant current drive, or at least some form of current limiting.

If you were to use an audio (linear) amplifier it would need a much higher bandwidth than 10 kHz if the pulse is not to be seriously distorted, 10kHz is only the fundamental frequency component of a 100 us pulse, and depending on how much degradation you can afford, I would think you need a frequency response to around 100 kHz. How frequent are these 100 us pulses?

An LED with current limiting and reverse voltage protection will present a seriously non-linear load to the amplifier which it may not handle very well.

As far as I can understand, you are trying to drive the LED with 100 us wide pulses, so the current will be either fully on or fully off, only two states. Therefore you do not need a linear (audio) amplifier, just a switched supply.

Have you considered better amplification between the Arduino and the LED, using higher gain transistors or an extra stage? Is the optocoupler necessary? Could you please post a circuit diagram and waveforms of what you are trying to do?


Looking at the data sheet of the IRF1405 I note that VGS(th) (the Gate Threshold Voltage) is between 2 and 4V, add to this the 0.7V Vbe drop of the 2N2222 and you could be barely driving the IRF1405 on with a 5V supply. And this assumes that the output of the Arduino goes all the way to Vcc (5V).

So why are you using an emitter follower configuration to drive the IRF1405? Ground the emitter of the 2N2222, add a suitable collector resistor (say about 1K) and drive the gate of the IRF1405 from its collector. OK, your signal will now be inverted, but a simple software tweak will fix that. In this configuration Ic of the 2N2222 will be 5mA and it has an hfe > 50 so a base current of 1mA is far more than will be needed to drive it into saturation and you can calculate the base resistor accordingly.

I can't find a value for VGS(max) for the IRF1405, but the data sheet gives characteristics up tp 10V, so the 2N2222 stage could be run from a Vcc of 10V (if available) in order to drive the IRF1405 harder.


simulate this circuit – Schematic created using CircuitLab

Above is a "back of an envelope" design. WARNING - It is untested and may contain errors!!

With the Arduino output at 0V, Q2 is turned off, V1 = 36V and hence Q1 is also off. V2 is pulled down to ground via R6 and M1 is off, no current flows through the LED.

When the Arduino output changes to +5V the base current of Q2 is limited to about 1mA by R1 and it turns on. The collector current will be limited to 2mA by R2 and Q1's Ib. V1 will be near 0V (Vce sat). The base current of Q1 will again be limited to about 1mA by R3. Q1 will be turned on and R5 and R6 set Ic at about 3mA and form a voltage divider defining V2, the gate voltage of M1, at 6.5V. Allowing a nominal 3V for Vgs, V3 will be 3.5V and therefore R4 will define the current, about 1A which flows through the LED. Note that R4 may need to be a high power resistor, under continuous load it will dissipate 3.6W. This will be reduced by the duty cycle. But expect magic smoke if you turn the LED on continuously with a lower wattage component!

  • \$\begingroup\$ Opto-coupler isn't really necessary, at first i thought some isolation would be nice since i will be switching relatively high voltages. I dont know any other form of power amplification other than transistor driven mosfet. Schematics with the best circuit i was able to come up with and the simulation results for led current added. I was able to get down to some acceptable 15us pulses by tweaking the resistors in the circuit, but in reality, will this just work like that? I also didnt post other results but currents and power dissipations were under limits for each component. \$\endgroup\$ Sep 6 '19 at 7:09
  • \$\begingroup\$ Yeah, as i also realized later, the problem was the VGS of the IRF1405. I also found that inverted configuration you proposed, on the internet. But if the arduino somehow delays pulling the output high at the startup, or transistor goes bad, it would turn the MOSFET on for much longer than a short pulse. I also plan to overdrive the LED later, maybe upto 10x-100x of rated current, so leaving the led on for more than a few microseconds would be catastrophic i think. Thats why i want the MOSFET switch in a normally closed configuration. \$\endgroup\$ Sep 6 '19 at 11:42
  • \$\begingroup\$ @ÖmerGezer Then you could probably afford 2 inverting stages (2 X 2N2222) to get the signal "the right way up". All sorts of possibilities present themselves with this setup. For example drive a 2N2222 from the Arduino, a 2N2907 (PNP equivalent to the 2N2222) as the second inverter stage with both transistors working from the 30V Vcc and suitably chosen resistors to limit base and collector currents and drive voltage (VGS). \$\endgroup\$ Sep 6 '19 at 12:28
  • \$\begingroup\$ That also makes sense, i currently have a seemingly working configuration i posted as an answer below, but i'll also keep the dual inverter in mind. Thanks! \$\endgroup\$ Sep 6 '19 at 12:35
  • \$\begingroup\$ "it has an hfe > 50 so a base current of 1mA is far more than will be needed to drive it into saturation " Nope. Saturation is noted for low gains. The general rule of thumb for driving a transistor into saturation is to assume a gain of 10 to 20, with 10 being the more widespread and conservative assumption. \$\endgroup\$ Sep 6 '19 at 13:35

An LED strobe driver at that power level needs some significant attention to managing current during the pulse period to avoid frying the LED, and a fast recovery time after the pulse. Neither attribute is in the realm of an audio amplifier, which by its nature will be bandwidth limited so will do poorly on both counts.

A switch type driver like you're prototyping is a start. But your driver isn't so well-controlled. It can't take advantage of being able to accurately pulse-drive the LED at high levels to achieve peak brightness.

I found a couple of ICs that can do this. I'm not recommending a specific product, but rather show these as examples of system designs for this problem class.

IC #1: a Linear Technology (now ADI) solution for a 3A pulsed LED intended for machine vision. https://www.analog.com/en/analog-dialogue/articles/led-driver-for-high-power-machine-vision-flash.html (datahseet: https://www.analog.com/media/en/technical-documentation/data-sheets/LT3932-3932-1.pdf)

This device will produce pulses down to the singles-of-microseconds range if you need that short of a flash.

IC #2: Here's a similar reference design from TI: http://www.ti.com/tool/TIDA-01081

  • \$\begingroup\$ Such an application specific ic would be perfect to use but i cant find any supplier in my country that holds these products in stock. I'd prefer to do the job with commonly available components. \$\endgroup\$ Sep 6 '19 at 7:07
  • \$\begingroup\$ @ÖmerGezer " i cant find any supplier in my country that holds these products in stock" Nonsense! LT3932 is in stock with Mouser and Digikey and they both ship worldwide. Eval board too. \$\endgroup\$
    – winny
    Sep 6 '19 at 8:24
  • \$\begingroup\$ @winny , they ship worldwide, yes, but why wait 2 weeks if it's something that can be accomplished with components i can get from stores that i can simply walk in? Also being an ameteur, using discrete components would give me a better understanding of the basics. \$\endgroup\$ Sep 6 '19 at 8:42
  • \$\begingroup\$ You will have it overnight or in two days max. Are you in school, then yes by all means. If you work at a company then time is of the essence. \$\endgroup\$
    – winny
    Sep 6 '19 at 9:19

After researching a bit more, and trying to improve my MOSFET switch circuit, I realized that the MOSFET I used before (IRF series) isn't able to switch properly with 5 volts coming from the transistor. So I found a spice model for IRL540, replaced it with the IRF one. And also instead of the pull down resistor, I added a PNP transistor to pull the MOSFET gate down fast enough. I can now easily get 0.5 us pulses at the LED module according to simulation.

I might open another question related to this NPN/PNP bridge driven MOSFET switch if anything goes wrong in the physical circuit.

Thank you all!

enter image description here enter image description here


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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