Paralleling LEDs is OK ?

[eem2am]

Hello

The person called "Sunnyskyguy" in the following thread is (very academically) arguing that putting power LEDs in parallel with absolutely no means of current equalisation is absolutely fine................

http://www.edaboard.com/thread253416.html

-Please can you help to debate against "Sunnyskyguy"?

"Sunnyskyguy" also trys to make the same argument in this post....

http://www.edaboard.com/thread254194.html

 

[Harald Kapp]

I'm sorry, but "Sunnyskyguy" is not absolutely wrong.
While what he's doing may not be the best engineering practice, it can be done.

Contrary to many of the discussion here in this forum, the main difference is that "Sunnyskyguy" is talking about driving the LEDs from a constant current source, not a voltage source. So the first part of saving an LED's live is ensured: current regulation.

Now the tricky part is how to divide the currrent from one source equally between more than one LED in parallel. Using resistors is a proven technique but leads to additional losses. As long as the characteristics of the LEDs match reasonably well, you may parallel them without additional resistors.

Assume you have just 2 LEDs where on has a slightly smaller Vf at a given current than the other LED has. Now look at the characteristic of the LED in terms If=f(Vf). Since the voltage across the two LEDs has to be the same, this will lead to an unbalanced current distribution between the two LEDs with the additional condition taht if1+if2=Isource. In fact, if the characteristics were very different from each other (e.g. paralleling a red LED with a blue LED) this will lead to one LED (red in the example) taking all the current and probably being destroyed.
This reasoning is easily extended to more than 2 LEDs.

So "Sunnyskyguy"'s argument is valid (only) for closely matching LEDs and he's gaining efficiency by sacrificing probably a bit of the LEDs' lifetime.

Harald

 

[KrisBlueNZ]

Harald, is thermal runaway a factor with high power LEDs? If Vf drops as temperature rises, and the LEDs are (at least partly) thermally isolated from each other, the LED that starts by drawing the most current will draw progressively more and more as it gets hotter, and eventually end up drawing all the current, and only a small initial imbalance would be enough to cause this problem.

 

[(*steve*)]

I've read his posts in some detail.

His argument -- as I understand it -- is that the incremental resistance (he calls it ESR) if the LEDs in series adds up to a substantial value and can replace the series resistor that you might use for improving current sharing between strings of LEDs.

I note that he recommends this only when used with a constant current source, not with constant voltage.

It also appears to require that you have reasonably long strings of LEDs, more so if they are high power LEDs.

He also recommends LEDs from the same batch and bin, both of which may be harder to achieve for a nobbiest as opposed to someone buying in huge quantities.

They are interesting arguments, but I'd still like to see the datasheet or application note that describes this technically.

If someone is a member of that other forum, ask him on my behalf. I'm always keen to learn more. I understand the principles of what he's saying, but I suggest that there are still a number of important caveats.

 

[Harald Kapp]

Kris,
good thought. This aspect passed me when I was considering the matter.
Yes, thermal runaway may be a problem, depending on how well the LEDs are thermally coupled. This will also depend on how much power is provided to the LEDs in total. Is there any reserve planned into the scheme or are the LEDs driven to the limit? We don't know.

As I stated, I do not consider this to be good engineering practice - but then again the definition of "good engineering" depends on the goal. If the goal is a "cheap product" not reliability, it might be o.k. to leave out the resistors. On the other hand, building a "cheap product" conflicts with the requirement for selecting the LEDs (same batch, same bin) which adds to the cost of the LEDs compared to non-selected LEDs.

And then there is the matter of the different brightness due to unequal currents..

Personally I would add the resistors.

Harald