Vf variance and parallel LEDs

[MrNaz]

I am thinking of building a light with LEDs in 2P2S array, something I swore I'd never do. I am paranoid about Vf variances, and driving LEDs unevenly when putting them in parallel. I want to ensure that a driver that is pushing a given current results in the current being evenly divided between the parallel paths of the circuit, and this requires the two LEDs to have the same Vf values. So, I need to bin my emitters by Vf.

When binning my emitters by Vf, can I assume that the Vf at some sample voltage, say, 350mA, will be a good indication of the Vf at other current levels? I.e., can I assume that two LEDs that require the same Vf at 350mA will also require the same Vf to push 1000mA?

 

[HarryN]

Of course you know the challenges already, so its just a matter of fine tuning the binning process.

In a perfect world, you would run the LEDs at the planned drive current for 24 hours to stabilize the Vf prior to the binning process. I can see your face from here already.

There have been a few posts in this section where builders have applied "electroshock therapy" to LEDs to drive down the Vf. More or less, they did very short, high current bursts to the LED and it seemed to have similar effects to a burn-in. I would not call it an ideal technique, but it did seem to work on the SST type leds.

Some people put a small resistance into each leg of the 2S2P arrangement to help balance out the current. In theory, you could get fancy and instead use thermisters that increase in resistance with temperature. This could help neutralize the reverse tendency of LEDs.

You didn't mention which brand of LEDs you are using, but frankly, some brands are much better than others for Vf consistency. My experience is that Lumileds has probably the best Vf consistency followed closely by Cree.

It is useful to note the failure modes of modern LEDs is somewhat different as well, in the rare event it should happen. A Rebel is now designed to fail as a "short" while most other LEDs will fail as an "open".