Another Voltage Drop Thread - So Called "Fairy Lights

We’ve been using Fairy Lights for some time now, now in addressable form, these from BTF -

With a more than ample power supply ( 5V, 30 A) there is considerable voltage drop with even one 50 px string. I’m measuring 1.3 volts drop at the first string.
The result of no additional power injection to a 250 px series string is as expected -

The (magnet) wire on these string is attractive from an aesthetic standpoint as it nearly disappears from view, ( its insulated diameter is 0.013 in) the downsides being fragility and, apparently, voltage drop.

Having to inject power every other string pretty much defeats the aesthetic purpose of “disappearing wires” and makes wiring up larger arrays cumbersome and unattractive with junctions in the field…

So, I guess the question is, granted other discussions about LEDs here, are there any good alternatives that any of you-all can think of that would satisfy the requirements:

Low wire visibility, large beam angle, high brightness and color accuracy (of course). Longer strings of 100 to 200 px, not needing power injection mid-string would be great. 12 volt would be fine . Maybe even 24 volt.

I recall talking with the creators of Radialumia at TTID one year, and they wired up mains with 24 volts, as I recall -

BTW this is a pretty interesting watch -

And that is a way-ambitious project. A beautiful install nonetheless.

Sure appreciate any input you may have.

Have a well-lit day. Cheers.

To be clear though, you could easily inject power every ~50 pixels (let’s roughly call it four lengths, two up and two down…)

All you have to do is add common voltage and ground lines at the top or bottom. If you need to add another power supply (I doubt it, depends on how big your single one is), easy to do that as well.

But yes

12v has more oomph and 24v even more, which is why longer runs can happen, but for 5v, you might be able to get 100-200 pixels with better wire, but injection around there is still recommended.

The voltage drop is based on current.

For strings like this I would avoid full brightness solid colors. Patterns with varying brightness will go a long way. Of you do want solid white, drop brightness to reduce current and lower voltage drop.

1 Like

Thank you @Scruffynerf and @wizard for your helpful comments.

We are just trying to get as close to “no apparent wires” e.g. typical vinyl-coated, as possible, in string format, also avoiding the “porcupine” look of many strings.

We may indeed end up using 100 led strings with power injection at both ends.

Otherwise, wondering about the following, available in 100 and 200 led strings, I think @jeff mentioned wiring them up in a volumetric cube - They get mixed reliability and durability reviews - wondering about voltage drop in a 200 led string…

Actually that was me… And those are weird, but yes, they work well in 200 led strands, power on one end.

I haven’t checked the voltage drop, BUT you can’t combine serially. You could run them individually, or with a output expander.

I had good results running my 200 LED fairy string at 100% brightness for patterns that didn’t tend to light up every LED at the same time. For other patterns I reduced the brightness as it was passed to hsv() until I didn’t see any color shift.

I noticed the patterns the controller it shipped with did similar. The whites were never 100%, and most patterns were sparse.

Of course you can inject power to drive higher power patterns, but then you have to run thicker wires. If you are going around a frame, you could tap in where it wasn’t super noticeable.

1 Like

I measured voltage drop behavior for the “Fairy Light” 2812B’s in question. Looked at injection at various points, at 200 and 250 px strings, and at just running a return ground to the string end instead of full injection.

Optimal results were achieved with injection at 150 px in a 200 px string. For 250 px double injection or single injection with the added ground line does the trick.

Here are the tabulated results -

Volts Drop 1.pdf (104.7 KB)


I have several sets of those, and knew that the voltage dropped off pretty quickly, but it’s awesome to see it quantified like that. Thanks, @JustPete!

Indeed it does. And injection is best, well ,where it is best, not necessarily at the string end.

Here is a more easily, and fully readable file (adjusted the color to be read without blanking).
VoltsDrop2.pdf (110.0 KB)

Curious the ground-limited behavior - just running a ground from the string end helps substantially.

Note: I am tying to limit the size of the wiring bundle for this setup.

Now I’m thinking about grounding this DC end of things in general, specifically WRT static buildup. And, about whether there is a better, more robust alternative out there that gives the "near 360-degree look)

@Scruffy do you have any wisdom to impart, or @wizard, oh wizened ones? It seems this LED arena is none too simple, Re your discussion with others noted above, and in this thread and with @wizard in this thread - Timing of a Cheap Strand.

As such, I remain in your debt.

Happy Fourth!

@JustPete ,
You hit on an interesting subject. These bus addressable LEDs share a data line through the entire strand. Power and ground are also shared. Something interesting happens when voltage drop occurs. On a normal addressable LED, the data is regenerated at each pixel. So if you had a pixel that was running at 4 volts it would output data at 4 volts. With these bus addressable LEDs the data is still at 5 volts from Pixel blaze, even if the last LED is down at three or four volts.

Voltage drop isn’t just about the positive rail, the ground rail will rise as well. So a 0 to 5 volt signal might end up looking like a negative one to four volt signal at an LED with 2 volts dropped. So tying even just the ground at the very end of the strand would help out quite a bit by bringing ground back to zero.

To be clear, the leds he’s using aren’t the bus addressable ones, they are the “3.5” wires which do have data in and out, and alternate that fourth wire. (So yes, they regenerate the data)

That said…

@JustPete, curious if you get similar results from

As it would be a good validation of that calculator’s prediction accuracy

Well, its an interesting model -

First to convert from kcmil to Ga -
[](https://Wire size conversion calculator)

If I put in all the relevant data: 200px; 5 volts; 28 Ga wire (0.160 kcmil); injection at 0 and 150 px with a short run… The model yields some interesting data but the results are controlled almost fully by assumed “intensity”. All 100% (full white I assume) tanks the system according to the model, giving negative 3V at about px #72.

20% intensity in the model yields similar results as actually measured with 5V injection at px 0 and px 150. Recall that I ran the system at full intensity, blue.

I measured wire diameter at 0.0135", corresponding pretty closely to 28 AWG with an assumed insulation coating of about 1/2 mil. Unfortunately, the manufacturer, BTF claimed that they are 22 AWG.

Final note for now, this “series” feed of the data seems just so archaic to me. Like old-style Christmas lights. But I guess its in the interest of low cost and minimizing material. But sheesh, one pixel goes out and there go the rest!

First, my apologies, I had mixed up your led types with the bus addressable type. Chained LEDs don’t have nearly the same relative voltage offset issue, but it can happen for long power runs to the first LED or otherwise in between long gaps where resistance still drops voltage across power lines.

They do make LEDs with backup data, so that any single contiguous issue doesn’t break the rest, but these are naturally more expensive and require more wires. I haven’t seen these on fairy wire style LEDs yet.

The “bus-addressable” LEDs shouldn’t have that problem, since they are all in parallel. A bad LED would not likely impact the bussed data. The downside is you can’t splice them like chained ones since they have an address programmed/burned into each. Still, these can come in up to 200 pixels, and can simplify some wiring since you don’t need to chain them to keep unique addressability (but are limited to 200 per channel).

That might work great as I am now running five channels at 200 px each.
Practically speaking I am now pretty freaked-out about the delicate nature of these fairy lights.
The “artistic challenge” is to achieve a similar effect as shown in this video - WS2811B Fairy Light Vis

Any recommendations for specific alternatives are most welcome. I am certainly considering more robust wiring.


There’s some more info about these on this forum, and a few other places as well. I’ll try to track down the others, here is the thread where it started:

In my opinion, these have slightly better diffusion and cast a more even light in all directions. Some people have also reported that these are more resilient to physical stress. It makes sense, these are made from 3 solid wires that are spaced evenly apart and don’t have trimmed wire nubs.

1 Like

Yeah, I’ve been very happy with the ones I’ve bought, and the wires are quite sturdy. There is a distinct “front” and “back”, but it does light up both side, though the front is brighter (of course)

I do recommend them, and the positives (including that one led failing doesn’t kill the rest, just that one dies) outweigh the negatives (the 200 pixel limit for one)

Thanks @wizard and @Scruffynerf, both . Just ordered some.

Is this

still a thing?

What thing? We still don’t know exactly how these work (they are clearly ws2812 compatible but aren’t ws2812/etc known chips, we don’t know what they are…)

@Scruffynerf Hmmm… mystery chips. What I meant was the first pixel won’t light thing.