On LED systems for live events & installations

Why distance sucks, but you can still shine.

This is an excerpt from a letter I wrote to a friend today, and as such, is in mixed voice and may contain minor mistakes or hanging threads. -5s

Over any significant distance, the major fights with LED systems become V-droop & data pollution/corruption. When we have long runs (longer than a couple linear feet) we need to think about these kinds of issue for two reasons that’s actually one reason; we have more copper. As runs get longer (either on PCB traces w LEDs or over copper wire to get from your driver to a structure) you add more and more material to be affected by all the physics that we are fighting against.

In power space, voltage falls off as a function of the growing impedance of line length, and doubly so as you load down the line with more and more LEDs. The end of the strip closest to the driver might read 5,12,24v, but the end might be 4.5,10,19v. This is V-Droop.

In data space, what we’ve managed to do by accident is build a crappy radio antenna. Actually, it’s not all that bad, and the longer it gets, the better it is. Anything moving through the air, and disturbance of the local magnetic field, and your data lines are going to pick up on it. You start having signal integrity issues. This is why * audio XLR is 3 pin- this is why Ethernet is 4 twisted pairs- you want better noise rejection. If you have to travel a long distance, you’re going to have to contend with noise.

So what do we do? Well, first is keep your runs as short as possible. “But I need 550’ of cable for this deployment” - then you better not be trying to run it over anything that’s not a differential pair, and god forbid it’s caring power.

For power, make your lines shorter and add more of them, do power injection, use higher voltage. Our system for crystal is 24v w current limiting resistors on each LED inside the bars - we essentially don’t have V-Droop issues ever, at least with that part of the system. The amount of vDroop is a function of line length, cable impedance, and driving voltage. Use lower impedance cable of higher gauge, use higher voltage, use shorter runs.

You should also split out your power and data runs any time you need to cross a large (where v-droop is significant) distance. When you do, you’ll want to add another power supply at the remote end. This will bite you in the ass if you don’t mind your Ps and Qs- you need to ground the second supply. Over the distance, you’ll need to carry data, data ground, and power ground. If they’re shared, two wires. This is because different PSUs will generate slightly different voltages, and untethered, will drift slightly over time, as they don’t necessarily have anything determining that this 12v and that 12v aren’t a volt different at both ends. Ground your supplies together, do not connect your power rails. You’ll damage shit, overload your PSU, or just consume a whole lot more power than you want to.

When you pull power away like this, all you get is a data line screaming across the void at khz - mhz frequencies. It’s a long, thin cable, and it picks up bullshit. A differential pair will help. Two data lines wrapped around each other, with an inverting amplifier at each end to phase cancel the noise. You can also do it passively by shielding one of them. Use cable with thick, braided shielding for long runs. You will have a much better time. Data pollution looks like flickers- LEDs juuuust not quite doing what they’re supposed to. It devolves into a sea of colorful chaos if you keep going. If you run it far enough on the wrong impedance cable , your signal will degrade to the point that it can’t be read by the LEDs- all the xxx-going transients will have been filtered away. So!! In practice what do we do????

Well Test that shit!! You’re gonna put down 150’ of cable ? Try it in the shop. And don’t just do it in a coil, toss it out. Inductive forces on crappy unshielded cable will muddy your readings. Use multiple PSUs, one at each device if you’re working with 5v. With 12, you might be able to get away with longer runs. With 24, we run [one of my projects] with no power injection at all [with more than 4k LEDs over 200' of stage]. Attend to your data streams. If you need to go an absurd distance, consider just having a totally separate system. A lot of very clever engineers have spent decades making IP packets go loooong distances over Ethernet. Buy a net switch and a second driver, drive two endpoints instead of one, and crossing a football field will suck a lot less. Or just use one, and understand that there’s a limit- pray you don’t hit it in the field, and test in the shop.

Keep good notes. future you (and me and any other techs) will thank you.

Blow shit up on the bench not in the field. Make prototypes, test ur shit.

Ask questions. Most engineers made the mistake of having engineering as a special interest, and will talk for hours if given the chance.

-5s