I'm about to build a custom high-power lamp. The lamp will operate a total of 60 high-power LEDs (operating point: 2800mA @3.1V per led) connected in series.

The original plan was to have the LEDs driven by a high-power constant-current LED driver, saving complexity in the circuit and ensuring that - since the same current is passed through each LED - all of them will be lit with the same intensity. Moreover, the constant-current driver also acts as a protection against short-circuit failure of the LEDs, because would any of them fail short it will adapt its terminal voltage to achieve the set current automatically, avoiding thermal runaway of the whole string.

However, what I'm not able to do is to identify a suitable protection for the LEDs failing open. Indeed, since there are many LEDs in the array and they are connected in series, I'd like the array to keep working even if some of them fails open.

I surveyed the market but found no component to achieve LED open protection. Indeed, most components designed for that purpose can't operate at such high values of current (2800 mA) and the same applies for the majority of Zener diodes I found. Therefore, my idea was to employ a Zener nodes + a NPN transistor in parallel with the LED.

The principle of operations should be that, once the LED fails open, the current regulator senses a high resistance and adjusted the terminal ∆V according. Once the voltage drop across the Zener becomes higher than its breakdown voltage, it allows a current flowing to the base of the transistor, opening up the collector-emitter line. This way the Zener diode is dissipating much less power, and most current is carried by the transistor (conversely, high-current transistor are easily found on the market).

I tried this circuit on Falstad and it seems to work; however, I'd like to have other opinions on this application because I'm worried about the model used in Falstad for the Zener/NPN Transistor being too simplistic, making me missing something I shouldn't.

Moreover, do I need additional resistances in the sub-circuit despite working in constant-current?


simulate this circuit – Schematic created using CircuitLab

  • 1
    \$\begingroup\$ You probably wont find this in any textbooks, but zeners reliably fail to short, so you could just put a zener across each LED, if the LED fails to open circuit then the zener will quickly overheat and fail to short. Outlandish as it may seem, conventional fuses operate by deliberately destroying themselves so this seems like a logical corollary. \$\endgroup\$
    – Frog
    Commented Dec 26, 2021 at 20:22
  • \$\begingroup\$ This seems a smart solution; a brief search on the internet confirmed that most Zeners fail short. I will surely consider it. \$\endgroup\$ Commented Dec 26, 2021 at 20:39
  • \$\begingroup\$ Actually I’d expected some fierce pushback from the community on that one ;-) \$\endgroup\$
    – Frog
    Commented Dec 27, 2021 at 6:43

2 Answers 2


As mentioned in the comments, an antifuse is an option. The voltage can rise momentarily (up to whatever the source's limit is) and then the antifuse fails shorted.

A solution might not be needed at all, however. If the CC supply is under design as well, then simply design it to deliver a maximum voltage nominally above the maximum string voltage.

The biggest hazard, for this number of LEDs, I suppose is the hazardous voltage across the whole string. The voltage across a break could rise enough to arc over, and then cascade failure might occur (burned LEDs, arcing to the substrate, etc.). This can be avoided by using a lower series count, and wiring strings in parallel (with an optional current-sharing resistors for each string; how much, or if at all, depends on Vf tolerance of the LEDs used, desired degree of matching at operating current, and nominal operation versus maximum ratings).

(Note that you do not wire LEDs in parallel first, and then series: LEDs won't share current well in each parallel tier; doing strings first, gives more chance for errors to cancel out. Even for linear resistors, the statistics are better for series-parallel than parallel-series (which skews low), but it's exacerbated by the nonlinearity, the low incremental resistance of the diodes.)

Alternately, if an off-the-shelf CC supply shall be selected, simply design the LED array for the nominal-max compliance voltage and LED string voltage. Select one with a modest output voltage (say 30-70V?) and enough current for the total strings.

The parallel-strings strategy does provide a degree of redundancy, in that if one string opens, the others take the load; the array could be designed with some derating in mind, to avoid overpowering the remaining strings. Else, a protection circuit might be desirable to account for this possibility, monitoring the current on all strings for example.


On the face of it what you propose is valid, but it does add significant cost and complexity. I’ve designed a number of commercial LED products and don’t recall ever seeing a LED fail to open circuit though. The 2N3904 that you’ve specified won’t handle the current or power, as you’re probably aware; you’ll need something that can handle more than the 10W rating if the LED, as the zener voltage + Vbe drop must necessarily be somewhat higher than the LED forward voltage.

  • \$\begingroup\$ Yes, I just picked the first NPN in CircuitLab and forgot to edit the labels. Thanks for pointing out that the transistor must handle more than 10W! I totally missed that the voltage would be higher. \$\endgroup\$ Commented Dec 26, 2021 at 20:30

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.