Here is a circuit that seems to work pretty well for allowing a voltage to control a current of up to maybe two amps.


It works on the falstad simulator (the triangle wave voltage input produces an identical current waveform thru the diode which is the LED -- Not only does PWM control work, setting input voltage to half will actually halve the current) and I will be breadboarding it with my own components but I want to figure out some things before I can finalize the circuit design.

The waveform at the left represents the voltage that regulates the current through the diode. The triangle wave (the 40Hz and the choice of triangle wave are arbitrary) is at 5V. It's connected to the op-amp through a voltage divider, and the pink resistor on the bottom right is chosen in coordination with the voltage going into the opamp (after the divider it is 100mV). The purpose of setting it to \$\frac{1}{7}\Omega\$ is to minimize heat generation. It is at that value because I wanted 700mA through the diode.

The DC voltage source on the right is 10V.

Now I looked at the spec sheet for my opamp, the MCP600x and it's absolute max current is 2mA. The Darlington pair transistor I am using (BD681) has a \$h_{FE}\$ of 750 which means that when I reach this limit the collector current is \$2mA*750=1.5A\$. This means I shouldn't use this circuit to drive a 3 amp LED. Do they make rail-to-rail opamps with higher current ratings? My transistor is rated to 4A (though it will need some heatsinking since it's got a 1.3v drop)

Since the transistor dissipates quite a bit of heat at high currents, I think a MOSFET would be necessary to reduce inefficiencies. However I haven't had very much luck getting a MOSFET circuit to have good current regulation because there is no conductive path from the gate to the source. I am able to use it as a switch, however, so PWM operation is okay.

I guess my question is: is there a relatively simple way to build a mosfet circuit that functions like this one, but capable of more current (and higher efficiency)? Also, which type of MOSFET should I be using (p-type, n-type... other types?)?

  • 1
    \$\begingroup\$ Why not use a "Trarlington?" Use a third transistor to make a triple Darlington - that should reduce the base current to a range the op amp can handle. \$\endgroup\$ – Bitrex Jun 24 '11 at 14:01
  • \$\begingroup\$ That is a very interesting idea. Certainly nothing stops me from further amplifying the gain in that fashion. Thanks! \$\endgroup\$ – Steven Lu Jun 24 '11 at 14:09
  • \$\begingroup\$ @Bitrex, Steven - aka "superdarlington" \$\endgroup\$ – stevenvh Jun 24 '11 at 14:21
  • \$\begingroup\$ @Bitrex: +1 for "Trarlington" although strictly a triple-Darlington would be a "Hexlington" :) \$\endgroup\$ – MikeJ-UK Jun 24 '11 at 14:32
  • \$\begingroup\$ @Mike - 1pA drive current? \$\endgroup\$ – stevenvh Jun 24 '11 at 14:46

This circuit uses a FET to drive a transistor, but you can replace the transistor with a darlington, so that you get kind of an hybrid superdarlington.

enter image description here

This should solve the opamp output current limitation.


Power MOSFET have another problem : they have significant gate capacitance, and with 2ma output current, it will be recharged very slooooow.... High speed drivers pump up to 1A of current just to drive gates...

So, I would first try to use low-power Darlington pair AND power transistor on top of it. This will give you enough of amplification.

  • \$\begingroup\$ Thanks for pointing this out! I doubt the simulator would take this into account \$\endgroup\$ – Steven Lu Jun 24 '11 at 14:12
  • \$\begingroup\$ Well, decent simulators should take this into account. Just add 1-10kOhm resistor in series with gate and you'll see. \$\endgroup\$ – BarsMonster Jun 24 '11 at 15:36

You can replace the NPN bipolar transistor in your circuit almost directly with a low voltage N channel FET. As long as the opamp has enough voltage capability to drive the gate, it will set the gate voltage to whatever is necessary to get the desired voltage (and therefore the desired current) accross the current sensing resistor. If you are using a MCP600x, then it's output voltage will be limited, and the FET needs to be a "logical level" type. That means it will operate over its full range with 5V on the gate instead of the usual 12-15V.

As for efficiency, linear control of the LED is a bad idea. Set things up to have just enough voltage to run the LED at the maximum current you want, then use PWM for lower currents. In other words, to get half brightness run the LED at full brightness half the time instead of half brightness all the time.

  • 1
    \$\begingroup\$ I understand that LEDs are best operated by using PWM but several of these new-fangled emitters can benefit from being driven at varying currents. The Cree XM-L for instance can provide an amazing 160 lumens per watt when driven at 350mA. It is also capable of almost 1000 lumens when driven at 3+ amps, but at a lower efficiency. If I want to set this emitter at ~30 percent of its max output, I would want to use a lower current rather than PWM at 3 amps. In fact I have an XM-L flashlight which uses PWM on its low modes. I reckon 300Hz pwm. And I think it would be better if driven directly. \$\endgroup\$ – Steven Lu Jun 24 '11 at 14:07
  • \$\begingroup\$ Also, what happened when I directly replaced the NPN with an n-channel MOSFET in the simulation was that the gate voltage became stuck on high and the MOSFET would not shut off. I had to fiddle with it and add a resistor to drain it to ground. It didn't give me linear control though. \$\endgroup\$ – Steven Lu Jun 24 '11 at 14:15
  • \$\begingroup\$ Bear in mind that if you are going to make an illuminator and view the scene with a video camera, using PWM can cause strobing/aliasing on the image. \$\endgroup\$ – MikeJ-UK Jun 24 '11 at 14:30
  • \$\begingroup\$ I'm glad you pointed that out, Mike. This is one of those things that are easily overlooked and adds to my list of reasons I like avoiding PWM (or at least coming up with workarounds for it) \$\endgroup\$ – Steven Lu Jun 25 '11 at 12:50

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.