I'm playing around with high-power LEDs (the "high" part is relative - 3W). I need 3 channels - 1 per RGB. Because I need independent control of the system, I would need either a multi-channel constant-current driver, or 3 separate single channel drivers. My question is - can I get away with a single channel driver, and use PWM dimming on the LEDs in offset phases? A (very) rough sample as I understand it would be:


simulate this circuit – Schematic created using CircuitLab

(Ignoring the lack of pull-up and gate resistors, and whatever other components I'm overlooking) - does this work conceptually?

  • \$\begingroup\$ Many constant current drivers use a low side shunt, so that wouldn't work. Do you know what you want to use yet? \$\endgroup\$
    – Wesley Lee
    Commented Dec 11, 2017 at 22:46
  • \$\begingroup\$ no, I'm still selecting, hence the question \$\endgroup\$
    – kolosy
    Commented Dec 11, 2017 at 22:47

3 Answers 3


Not enough gate voltage for an IRF530.

So about 1A per channel, not all that much. Why not just use three of these and a constant voltage supply such as 5V? If the LEDs are 3V-ish you'll have about 1V dropped across the resistor and 1V across the transistor (so about 1W each).

enter image description here

You can divide down the MCU outputs with resistors, or use resistors and an TL431 or LMV431 for each if you want to regulate the brightness independent of supply voltage.

Edit: Since you have a LiPo for the LED supply you have to call it a bit closer, depending on the LED and where you cut the battery off. The sense resistor can be reduced to 0.1 ohm or 0.05 ohm, and the MOSFET switched for a low- Rds(on) specified for 2.5V drive. That may or may not be enough- but you can get down to well under 100mV drop at 1A. If it is enough it may well be more efficient than a switching constant current supply. The op-amp is not particularly critical but it has to be a single-supply or RRIO type capable of working from the full range of battery voltages and without too much offset voltage compared to the sense voltage. There are a number of decent candidates.

  • \$\begingroup\$ never assume even LT is correct. The 1 ohm is too big and the FET part number is a controller not a fET and if this is a battery the Vgs(th must be < 0.6V with Vol for that op Amp +1 for idea -3 for errors \$\endgroup\$ Commented Dec 12, 2017 at 2:17
  • \$\begingroup\$ Mostly because I'm going to be driving this off of a single cell LiPo, which is relatively low. \$\endgroup\$
    – kolosy
    Commented Dec 12, 2017 at 2:44
  • \$\begingroup\$ @TonyStewart.EEsince'75 It's a correct number for a Vishay Siliconex MOSFET. The OP can adjust the parts if he likes the idea, or ask for help. It's intended as an example of a possible approach. If the LEDs are single dice there is probably enough margin. \$\endgroup\$ Commented Dec 12, 2017 at 3:41
  • \$\begingroup\$ @kolosy see edit above. Whether it's practical or not depends on the maximum LED Vf and the minimum cell voltage. \$\endgroup\$ Commented Dec 12, 2017 at 3:58
  • \$\begingroup\$ OK but this design wont work off a LiPo. The Voh, shunt R drop voltage unspecified Vgs(th)max will not even work at 4V as it was intended for 5V \$\endgroup\$ Commented Dec 12, 2017 at 4:33

The concern I would have with your idea is how the CC driver will respond to sudden changes in the load. If you break before make that may result in some extreme voltage transients in the output of the driver with rather unfortunate consequences.

Make before break may be less problematic, but again there will a sudden change in the voltage output which the driver and LEDs must recover from.

As such, I think I would shy away from this idea.

A better solution may be to use a boost regulator to buy you enough voltage to use some simpler current limiting circuitry.

Also be aware, you will not be able to gain the same brightness from a multiplexed solution.


Unfortunately the power LEDs lose efficacy, and are often limited to a ratio of 150% = Ipk/Idc_max current so that the thin wire bond does not fuse open or the current density on the chip is not excessive.

Thus operating the LEDs from 0 to 33% duty cycle at 3x the rated current will compromise the chip specs and lumen efficacy if you are expecting full performance.

Thermal design of 10 deg'C/W case temp or less is also critical for 3 Watt chips.

But operating at 0 to 33% duty cycle at 2x the rated current gives an average of 2/3 rated DC power which will work but at compromised brightness.

So it depends on your specs. Yes but sub-optimal.

A better approach is to choose a buck choke PWM regulation design for each LED and use a current shunt R to detect \$I_f\$ for each LED. A reverse clamp diode with a <50mohm RdsOn can work with a 0.5A low DCR SMD choke.

I presume that 3.7V means you intend to use a LiPo cell but that drops to 3V , so none of the linear solutions offered will work.


Let me give you another hint. This one has a range of 5000 steps on each color using PWM with inductive constant current regulation and a 50mV current sense.

enter image description here

  • \$\begingroup\$ that is not at all an answer to my question. \$\endgroup\$
    – kolosy
    Commented Dec 12, 2017 at 20:33
  • \$\begingroup\$ Well you never specified Vf vs If and Vbat min. in your previous question. Would you like me to do the complete design specs and solution? You failed to define the specs. SORRY a better answer only comes from a better question with specs. I promise to make more effort than you. But at least I explained why the shared CC will not work! Which is the correct answer? \$\endgroup\$ Commented Dec 12, 2017 at 20:57
  • \$\begingroup\$ I asked if a particular design approach works, not "how do you build an LED driver," or "does this specific implementation work." \$\endgroup\$
    – kolosy
    Commented Dec 13, 2017 at 3:47
  • \$\begingroup\$ Yes and I effectively answered No it does not work and why it compromises. \$\endgroup\$ Commented Dec 13, 2017 at 4:08
  • \$\begingroup\$ The answer you accepted however did not respond to your question , does it work? No you cannot regulate each independent when sharing constant current. Its a violation of Kirchoff’s KCL \$\endgroup\$ Commented Dec 14, 2017 at 7:54

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.