I have a small PCB that holds 6 LEDs (model, fV 3.2V, 30 mA), 2 parallel lines of 3 LEDs in series. The PCB has wires soldered onto it that go to a connector about 10cm down the line.

circuit showing 6-LED pcbs and connection points

In an installation with several of those PCBs, I've experienced that after a while a few of them only light up half the LEDs.

In the proper setup those LEDs are driven with a LED PWM chip. Outside the setup, i test them with a 9V power supply.

What I now want to understand is the behavior I observe when debugging those faulty boards:

  1. I test connect 9V in connection A (at connector) and only half the LEDs work
  2. Then I test connect 9V at connection B (directly on the solder joints on the PCB) and now all LEDs work
  3. I go back to connect the 9V in connection A (at connector) and suddenly all LEDs work from here as well

There is no visible sign of a broken solder joint or lose wire connection. (Obviously, there could be, as the boards are handsoldered. Also, most connection points are covered with conformal coating and I am now suspicious if the coating puts stress on some of the solder joints)

What makes LEDs sometimes turn on and sometimes not? Does this behavior indicate that either one or more of the LEDs are actually damaged, or is it more likely that there's a lose connection somewhere?

  • \$\begingroup\$ How are you limiting the current to the LEDs? \$\endgroup\$
    – HandyHowie
    Jun 10 '19 at 7:32
  • \$\begingroup\$ @HandyHowie With an LED PWM driver in the proper setup. But outside the setup, during debugging, I don't limit it. The LEDs have a 3.2V fV, powering 2 times 3 of them in series with 9V, i get ~37mA current, which corresponds to the LEDs current-vs-voltage graph from the datasheet. Therefore I am not worried of damaging the LEDs during testing. Should I be? \$\endgroup\$
    – evsc
    Jun 10 '19 at 7:46
  • 1
    \$\begingroup\$ TL;DR: Don't parallel LEDs. \$\endgroup\$
    – winny
    Jun 10 '19 at 7:58
  • 1
    \$\begingroup\$ ... and current limit under all situations. \$\endgroup\$
    – Andy aka
    Jun 10 '19 at 8:25

The problem you have is one that is a design problem. You need to realize that the three LEDs in one string may very well have a higher total voltage drop than the three LEDs in the parallel string. The ones with the lower voltage drop may light while the others do not.

Since you do not have any intentional current limiting scheme to control the current flow through the LEDs in your testing scenario you are depending upon the forward voltage drops across the LEDs to be more than the sourcing power supply. In this scenario the effective internal resistance of the LEDs and the resistance of the wiring harness are being used to limit the current. Operating in this mode the V / I characteristics of the LEDs can shift with time, aging of the LEDs and the current operating temperature of the diode junctions within each LED.

The solution is to provide for a separate driver for each string of three LEDs or to place a resistor in series with each of the three LEDs that are already in series.

  • \$\begingroup\$ interesting. you're saying that all manufactured LEDs have small variations in their voltage drop? Can't be large differences though. If one side has a summed up voltage drop of (3x3.2=) 9.6V, and the other let's say (3x3.0=) 9V, only one of them would light up? \$\endgroup\$
    – evsc
    Jun 10 '19 at 7:57
  • \$\begingroup\$ That is what I am saying although the difference may not be as large as the 0.6V in your example. This effect is often seen in the cheap LED flash lights that have 10 to 20 individual LEDs all connected in parallel. With a fresh battery all the LEDs will light but as the LEDs age and/or the battery capacity gets consumed some LEDs will flicker and some may just go off. \$\endgroup\$ Jun 10 '19 at 8:03
  • \$\begingroup\$ In the proper setup I use a current-limiting LED driver to power the LEDs. Is the different-fV-issue still a problem in that scenario? And, another question: So if I'd put a resistor (say 10ohm) on both lines, then both would light up, even if they have diff fV values? \$\endgroup\$
    – evsc
    Jun 10 '19 at 8:20
  • \$\begingroup\$ @evsc - Your "proper driver" may very well still have a problem with getting two parallel strings of LEDs to behave the way that you want them to. The resistor value you use you will have to determine what works in your scenario. I suggest that in your testing scenario that you will get better results with resistor values that are intentionally limiting the current through each string from a higher voltage source like 12V instead of operating in that no mans land just below the Vf total of the LED strings. \$\endgroup\$ Jun 10 '19 at 8:27
  • \$\begingroup\$ okay, super insightful, learning a lot here. thank you! \$\endgroup\$
    – evsc
    Jun 10 '19 at 8:38

I know this thread is getting old, but I just wanted to add a point for future readers. Diodes of all sorts often have an inverse Vf vs. temperature curve. Well, curve might be misleading since it's most often linear. In fact, diodes are sometimes used as temperature sensors for this reason. Knowing that, and that when you put current through a device, it self heats, I think you can see where this is going. For diodes, this means that the Vf is going to drop a bit after the device is on for a few milliseconds. This is the very reason that current limiting is so important when driving LEDs. If you have no limits, the LED is likely to continue heating up until it gives it's final flash and releases the magic smoke. BTW, the datasheet for the mentioned LED states that the Vf is nominally 3.2 but the max may be as high as 3.8, so there is potential for as much as a 1.8V variation in Vf in a stack of three LEDs (though it's highly unlikely). They don't bother with the lower limit, which could reveal a larger potential variation.

  • \$\begingroup\$ Interesting... The temperature dependence of the LED IV curve can be another reason for the current redistribution... \$\endgroup\$ Feb 18 at 21:39

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.