1
\$\begingroup\$

I'm looking at using a single constant current, DC-DC boost converter to power several (6 - 8) parallel strings of 3 LEDs, like the drawing below. This is powered by a single li-ion battery.

The example circuit without MOSFET switching

To control each string of LEDs, I'm proposing using a MOSFET like in the diagram below.

Possible way to control each string

Each MOSFET would be driven by a separate microcontroller pin either high or low.

As I understand it, provided the set current is high enough to drive all of the strings simultaneously this will work but might have some side effects:

  1. When disabling any string, the current will be shared by the remaining enabled strings, i.e. the remaining strings will increase in brightness. This means when only one string is enabled, it must be capable of being driven at the full current (8 x the current in the above diagram).

  2. If I used the MOSFETs to dim each string individually using PWM, it may be a significant source of EMI as these will be driven by non-slew rate controlled ON signals.

  3. Dimming using PWM would be less efficient than other methods but would offer better chromacity and intensity control.

  4. The LEDs used in the feedback circuit will need to be illuminated and can't be switched on / off, unless the whole DC-DC converter is disabled.

My questions are:

  1. Are my assumptions correct above?
  2. Are there any other disadvantages to this approach?
  3. Could the three LEDs used in the feedback circuit be substituted with a zener diode and resistor? The feedback voltage for this driver IC is ~100mV.

Any other thoughts on this, or other potential solutions would be appreciated.

\$\endgroup\$
  • 1
    \$\begingroup\$ You're relying on quite small resistors to keep the current balanced between strings, this small negative feedback may not be enough to overcome the positive feedback inherent in LEDs (as a diode warms up, its resistance drops). Perhaps more importantly, your feedback isn't monitoring the current through the additional chains, so you won't achieve constant-current operation. Unless your LEDs are (1) extremely well-matched and (2) thermally coupled, this won't behave itself, and it any case it won't behave the way you've mentally modeled it. \$\endgroup\$ – Ben Voigt Jan 21 '17 at 22:53
  • 1
    \$\begingroup\$ Interestingly, the first diagram comes from the datasheet of the DC-DC boost converter. With or without the MOSFET's, the feedback isn't monitoring the current through the additional chains. Using a larger ballast resistor would help. In favour of this possibly working is the fact that the LED's would come off the same reel. Are my other assumptions correct, ignoring the matching issue? I'd also like some constructive criticism on the Zener substitution instead of LED's in the feedback. \$\endgroup\$ – pythomatic Jan 21 '17 at 23:12
  • 1
    \$\begingroup\$ For example, you might want to consider ST's LED7708. \$\endgroup\$ – Ben Voigt Jan 22 '17 at 0:22
  • 1
    \$\begingroup\$ the design relies on the assumption that the leds have relatively small divergent forward voltage, or temperature coefficients. You will have to look up your leds and see if they differ much in those departments. for high power leds, they are more like resistors (of small value) and monitoring one string is good enough. \$\endgroup\$ – dannyf Jan 22 '17 at 1:15
  • 1
    \$\begingroup\$ if you want to monitor multiple strings, you can sum up the voltage at the feedback pin by connecting a resistor (1k for example) to that string's current sense resistor. Not perfect but close. \$\endgroup\$ – dannyf Jan 22 '17 at 1:16
1
\$\begingroup\$

You could try setting up an array of BJTs to act as a current mirror. That would balance your load currents in each leg. Even if the transistors aren't matched sets in the same die, they should be similar enough to use for this purpose since the 10 ohm resistor is providing feedback on VBE. The MOSFET RDSon must be well below 1 ohm to get good matching using this method.

schematic

simulate this circuit – Schematic created using CircuitLab

| improve this answer | |
\$\endgroup\$
1
\$\begingroup\$

As already said, string 1 is operated with constant current which sets the overall voltage for all other strings. So all other strings operate at a given voltage. Therefore assumption 1 is not correct.

2 - assuming the PWM frequency is in no higher than double digit kHZ range, it would likely produce very little EMI due to "fast" edge rate. For example, a "10Mhz" edge would naturally have much reduced power with a 10kHz repeating (PWM) rate. There are always exceptions, one would be that the circuit is somehow resonant to the high frequency.

3 - dimming using PWM can theoretically approach 100% dimming efficiency, what are the "other methods"?

4 - correct.

If the feedback string is substituted with a zener, then the voltage matching would get worse. For example, the temp coefficient differences between a LED and a zener would be worse than between 2 LEDs of the same type. And typical zener diodes are not very good voltage reference (comparing to any built-in reference of a voltage regulator for example), then you might as well go constant voltage power supply.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ With 3/, I was referring to analog dimming if a different topology was used - I believe that this would yield a higher power efficiency at the expense of possible colour shift with white LED's and additional non-linearity of brightness. \$\endgroup\$ – pythomatic Jan 22 '17 at 12:35
0
\$\begingroup\$

If you want to be able to turn off any or all of the strings, then a constant current drive simply does not work. As you've already discovered, if you turn off strings, then the current redistributes which will result in large variations of brightness.

I'd suggest you should drive the chains with a constant voltage. It's slightly less efficient, but may work for you.
All you need to do is make sure the voltage dropped across the series resistor is greater than the voltage change across all the LEDs (in one string) due to temperature changes. For example: If the voltage across the series resistor was say 2 V (the same as the LED's below) then you could turn off each LED string without impacting the current in other strings. Here the current is set at 20 mA.

schematic

simulate this circuit – Schematic created using CircuitLab

| improve this answer | |
\$\endgroup\$

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.