Recently I've come across two very different LEDs that have failed due to a soaking, but with interesting and almost identical failure behaviour. I'd like to understand what's going on inside.

Once dried both emit, but not as expected. At a reasonably high current (i.e. close to their maximum rating) they give a steady light which by eye is indistiguishable from normal behaviour. At lower drive currents the output is less than normal and flickers. The (apparent) Vf, fluctuates. At still lower currents, but enough to normal give clear emission, there's nothing.

The first is a 365 nm 1.2 W from LED Engin used in an aqueous photocatalysis by undergrads who presumably got it wet. This is run off a bench supply so I can test it more thoroughly. The second is a red 5050 single emitter in a rear bike light that got soaked in a storm. Failure is only apparent when on standlight mode (running with a lower drive current than if moving, powered by a 1F 5.5V capacitor that I initially blamed). A thorough drying made no visible difference (in fact dried, rinsed in DI water, dried, baked in a 100°C oven for an hour, tested at every stage).

This is mainly out of curiosity; I've got some background in semiconductor failure but nothing like this. The failure mode seems interesting in the way it manifests at low current by higher currents seem fine.

  • 1
    \$\begingroup\$ Might be dendritic growth due to electro-migration on the PCB/LED housing. Once those form a water wash won't get rid of them. \$\endgroup\$ – isdi Nov 8 '18 at 11:24
  • \$\begingroup\$ @isdi I've had the UV one under a microscope with noting visible on the outside before or after desoldering it. I may do the same on the red one \$\endgroup\$ – Chris H Nov 8 '18 at 11:54
  • \$\begingroup\$ Are all of your tests made with connections soldered directly to the LED? \$\endgroup\$ – WhatRoughBeast Nov 8 '18 at 12:23
  • 1
    \$\begingroup\$ LED cases are not hermetically sealed water tight. \$\endgroup\$ – Uwe Nov 8 '18 at 15:08
  • \$\begingroup\$ @Uwe yes, but what's happened inside, once the water has (presumably) been driven off? If it was wet and acting as a short I could understand it, but why flickering/fluctuating resistance? \$\endgroup\$ – Chris H Nov 8 '18 at 16:14

I think David's on the right track, but didn't explain thoroughly. Dendritic growths and other impurities form a parallel path to the LED. Now, these may not be linear, homogeneous paths...the impedance may change with time, temperature, humidity, etc. One possibility is that the heat from the LED helps to complete the path, so that it brightens and dims.

If you have current available that will keep the LED on even when the alternate path is saturated, it will appear steady. If the current can be easily absorbed by the other conductive path, it will likely flicker as the properties of that path change.


Just like you'd get by putting a resistor of the right value parallel to the diode.

  • 1
    \$\begingroup\$ Could you expand on how that would cause the flickering/fluctuating V_f? A resistor alone wouldn't seem to cause this, and I can't see where enough capacitance would come from -- but maybe I'm missing it. \$\endgroup\$ – Chris H Nov 16 '18 at 9:28
  • \$\begingroup\$ Typically, flickering is caused by an unxtable power supply. And particularly when you get the exact same flickering with different test cases. A small cap in parallel, that often fixes supply instability. \$\endgroup\$ – david Nov 17 '18 at 21:43
  • \$\begingroup\$ that was my first thought on the red, which I tested at home(and I still have my suspicions) but the UV was tested on a basic but reliable linear bench supply in work, where I have better kit. In particular I've put a bit of effort into characterising the constant current output of our bench supplies for LED spectroscopy, and this one is the steadiest \$\endgroup\$ – Chris H Nov 18 '18 at 7:51

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.