I'm trying to design an efficient circuit that is able to drive 4 independent (ie, connected in parallel) white LEDs and to control their intensity independently as well, through PWM. The thing is: most of the circuits I find were designed either for 3 "strings" (rows of LEDs), which is cool if you use RGB LEDs, or don't provide a way to control the intensity of each string.

Using multiple driver ICs, each one controlling a single row, would destroy the efficiency of the whole circuit.

The white LED I'm using: http://www.mouser.com/ds/2/311/LW%20Y1SG%20-%20Micro%20SIDELED-335725.pdf Forward voltage (max): 3.6V Current: 30mA

Also, two other requirements are: it should have low quiescent current (ideally < 1uA) and Vin(min) > 1.6V since the circuit may be powered by two alkaline batteries.

This is the first time I'm designing such circuit, so I'm very looking forward to hear some tips from those who have done this before. Thanks in advance!

  • 1
    \$\begingroup\$ Need... more... info.... "white LEDs" - that is very broad, the colour is meaningless, the voltage and current specs say a thousand words. What does "efficient" mean? 50%? 80%? 100%? \$\endgroup\$ – Tom Carpenter Dec 3 '15 at 2:43
  • 2
    \$\begingroup\$ I think your best bet is going to be to boost up to 4V to create a VCC for all the LED's, then control each one independently with a current sink circuit. You can PWM each current sink independently. Do you have a regulated IO output voltage available? \$\endgroup\$ – mkeith Dec 3 '15 at 3:10
  • \$\begingroup\$ @mkeith, yep that's probably what I'll use, thanks! I don't have a regulated IO output voltage, I was considering to use PWM directly connected to the transistor/switch. Do you think it's better to filter it in order to avoid current spikes in the LED? \$\endgroup\$ – rasgo Dec 3 '15 at 4:10

You could do a sort of modified high-side PWM controlled joule thief for each of the LEDs, with each having an independent PWM controlling the power though the LED.

Simple version might look something like this...


simulate this circuit – Schematic created using CircuitLab

This is different from a normal joule thief in that...

  1. The chopping is done via PWM rather than feedback. This lets you control the power though the LED.

  2. The LED is connected to the high side rather than ground. This means you will need more kick in the inductor to get the LED up to minimum voltage, but has the advantage that there is no leakage current though the LED when the PWM is off.

Note that there is not a straight relationship between PWM and brightness here. When PWM is off, LED is off and no current flows. As you increase PWM duty cycle, you will hit a minimum current needed to light the LED. Then as you increase more, you will get brighter and brighter LED until you hit a maximum. Past the maximum the LED will start getting dimmer and you will start killing the battery, so stop increasing the PWM duty cycle when you get to the maximum brightness.

This circuit should have very low quiescent current- basically whatever the leakage current of the transistor is.

  • \$\begingroup\$ This is a good idea. Better than my idea. Although there will be some sensitivity to V1 (which is can range from 3.2 to 1.8). \$\endgroup\$ – mkeith Dec 3 '15 at 4:22
  • 1
    \$\begingroup\$ If you are using an AVR to generate the PWM, then you could use the internal voltage sensor to adjust the PWM based on current battery voltage. There are similar hacks for other MCUs. I've done this and it works surprisingly well. You can connect to a variable power supply and fiddle with the voltage and the LED stays constant. \$\endgroup\$ – bigjosh Dec 3 '15 at 4:26
  • \$\begingroup\$ with VBATT sensing and compensation, I think this is a very good solution indeed. Great job! I have actually thought of doing this before, but never played with it. \$\endgroup\$ – mkeith Dec 3 '15 at 4:47
  • \$\begingroup\$ I have seen a few white LED's that incorporate silicon diodes to protect the LED from reverse bias. Such LED's could not be used in this fashion. Also, LED's in general cannot take very much reverse bias. I don't think that will be a problem in this circuit, though. \$\endgroup\$ – mkeith Dec 3 '15 at 4:51
  • 1
    \$\begingroup\$ @mkeith makes a great point about some LEDs not being able to tolerate a reverse bias. I'd never heard of this before, but it is possible that the OP's diode is one of these based on the fact that the datasheet does not give a safe maximum reverse voltage. In this case, the LED could be connected to the low side instead with maybe a MOSFET to turn off power to all LEDs when in low current sleep mode. Thanks! \$\endgroup\$ – bigjosh Dec 3 '15 at 5:04

Here is my idea. Consider this a sketch or snippet.


simulate this circuit – Schematic created using CircuitLab

VCC_LED should be the output of a boost regulator that produces something like 4V. L1, C1, D1 and M1 are meant to represent the boost regulator. Values are just guesses.

D2, Q1, R1, D3 and D4 are the current mode drive for ONE LED. That will have to be duplicated 3 more times for the rest of your LED's. D3 and D4 could be combined in a single small package. To set the current, adjust R1. Basically, one of the diodes approximately cancels out Vbe of Q1. So the voltage across R1 will be around 1 diode drop. Q1 will be in the linear, high beta region. This is not a saturated switching application.

I am pretty sure this will work with a minimum of fine-tuning. I have used circuits like this before in mass production. If PWM1 was from a regulated supply, D3 and D4 could be replaced with a single resistor. That is why I asked about that.


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.