Amazing Analysis Of A 350,000 LED Airport Art Project

Before you zip to the comments to scream “not a hack,” watch a few minutes of this teardown video. This 48 minute detailed walkthrough of a one-off art piece shows every aspect of the project: every requirement, design decision, implementation challenge, and mistake. Some notable details:

  • PCBs that are 1 meter wide (all one piece!)
  • 350,000 white LEDs
  • Carbon fiber enclosures
  • 1-wire serial bus (like the WS2812 only not quite) with 12 bit resolution (TLC5973)
  • Customized cable test jigs, PCB test jigs, and test modes
  • An exploration on ESD issues in production

It’s not often that one sees teardowns of professional projects like this, and there’s quite a bit to learn from in here, besides it being a beautiful piece of art. See more about the Caviar House “Emergence” project at the Heathrow Airport, along with stunning pictures and video of the display in action.

If you’re thinking about how you’d control 350,000 individual LEDs with 12 bit grayscale and have it look smooth, check out the processor requirements behind the megascroller, which only handles 98,000 LEDs. More recently, we asked how many LEDs are too many, and the answer was quite a bit lower than 350k.

29 thoughts on “Amazing Analysis Of A 350,000 LED Airport Art Project

    1. Doubtful. You would need to have some way to anchor them in place, they would need to be electrically connected together, and after all of that it would be encapsulated in resin. With a huge PCB like this just fab, pick and place, reflow, and cap it with resin. Then you drop one unit onto each side of the carbon fiber and for every arc and you’re done.

      1. I’m doubtful there are many pick & place machines that do 2m long PCBs. Our machines at work only do up to 450mm wide, and less deep. PCB manufacture may also be a challenge.
        The PCB size is impressive though, but surely something to avoid designing if it can be modularised.

        1. The PCBs are 1m long – 2 on each side of each arc.
          The place that assembled them can do up to 1.6m
          Though their machines are designed for long boards, it is possible to do long PCBs in multiple sections with a suitably configured conveyor system. Paste printing can be more of an issue though.

    2. Not really – electrical and mechanical interconnect would be a pain – you’d need to align them accurately to avoid visible discontinuities – encapsulating is enough of a PITA without adding to it.
      It would make no sense to do it as small as 10cm,though sticking below about 45cm gives you a bigger choice of PCB suppliers and assembers.
      Big PCBs aren’t inherently more expensive per unit area, it’s just you have less scope to shop around.

      1. Maintenance is also much more of a task with smaller PCBs. The bigger they are the easier they are to get working if something fails. I think I read that they are also emergency lights? so no matter what it does artistically it must function and be repairable.

  1. Couldn’t they just process and timestamp the data with a PC and control the LEDs locally in groups, using simple microcontrollers? Once you sync them, and buffer the data you can live with arbitrary latency. Unlike megascroller, which works as a display for an interactive device.

      1. LED tape is usually soldered together from ~1m strips.
        1m LED strips are pretty common nowadays. Even Neoden have a <$10K machine that can do them ( in sections using conveyor feed)

      1. So what bit of this, exactly, qualifies it as not a hack?

        Using LED drivers for 3x greyscale rather than RGB – hack
        Offloading rendering onto multiple peripheral microcontrollers to reduce communication bandwidth? Hack.
        Bending 2 SPI ports at 3.3v into 4 5v single wire ports? Hack
        Mapping 3x 10bit outputs into 32 bit SPI, then using a lookup table to reduce on-chip instructions and increase speed? Hack.
        Finding your resin changes the colour temperature of your LEDs? Hack.
        Troubleshooting board failure after encapsulating? Hack.
        Finding it due to ESD in an electrically isolated environment, then proving it with aluminium tape? Hack.
        Powering a neon bulb with said tape? Hack.
        Minimising the hole in the structure to insert a data splitting PCB? Hack.
        Multiplexing using one end of the long with a ceramic resonator to reduce timing errors on communication without pushing the size? Hack.
        Parada and light pipe status indicator, that self blanks after a couple thousand packets? Hack.
        Using a nut, a spare JST socket and a magnet to drag cable through partially occluded carbon fibre spars? Hack.
        stripboard testing board? Hack.
        Jigging timing to enable firmware updates with only one-way communications? Hack
        Getting that timing by padding the firmware with extra bytes? Hack.
        Using the output arcs to show the slow NOR flash erase time? Hack.
        A seemingly kinetic structure with no moving parts? Hack.

        Seriously, you can’t see the trees for the trolls around here. I’m glad I watched this, and it’s a pity you PITAs either didn’t or are so knowledgable that you wasted your time on a site like this to comment.

        1. “Using a nut, a spare JST socket and a magnet to drag cable through partially occluded carbon fibre spars? Hack.”

          That was it for me!
          After all the very clever “hacking” he had to do along all of the creating process, at every steps, he still had to find a smart way to pass cables through the frame.Come on!

          I’ve discovered mikeselectricsstuf a few month ago, was amazed by his videos, but I always wondered what he was doing for a living besides showing awesome stuff on youtube.
          Needless to say, I wasn’t disappointed.

      2. Whether or not you think it was a hack (I think it was), I thoroughly enjoyed it, and I wouldn’t normally watch a video this long.

        How about for next year’s April’s Fools – rename the site to Not-A-Hackaday and have articles like plugging an LED into a breadboard with an Arduino … and inserting batteries correctly to have things work.

  2. Not a hack, not an analysis, not a teardown.

    [Mike] gives us a great behind-the-screens look at how he put the electronics of this art piece together, and shares some insights on solutions to problems that you might not think of if you’re just playing around with Arduinos after school.

    Absolutely brilliant!

  3. Awesome project and very nicely done! What was the timescale you were talking about. I think you only mentioned it was “fairly short” but how many weeks did it take from planning to the finished installation? Could you tell about the individual steps like 2 weeks planning, 6 weeks manufacturing etc.?
    Again, nice project and I liked the magnet solution ;)

  4. From memory it was something like 6 months start to finish, and Christmas and Chinese NY was included in that.For a project that big, that’s borderline insane.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.