Fail Of The Week: Bright Idea For LED Signs Goes Bad

Typically when we select a project for “Fail to the Week” honors, it’s because something went wrong with the technology of the project. But the tech of [Leo Fernekes]’ innovative LED sign system was never the problem; it was the realities of scaling up to production as well as the broken patent process that put a nail in this promising project’s coffin, which [Leo] sums up succinctly as “The Inventor’s Paradox” in the video below.

The idea [Leo] had a few years back was pretty smart. He noticed that there was no middle ground between cheap, pre-made LED signs and expensive programmable signboards, so he sought to fill the gap. The result was an ingenious “LED pin”, a tiny module with an RGB LED and a microcontroller along with a small number of support components. The big idea is that each pin would store its own part of a display-wide animation in flash memory. Each pin has two terminals that connect to metal cladding on either side of the board they attach to. These two conductors supply not only power but synchronization for all the pins with a low-frequency square wave. [Leo]’s method for programming the animations — using a light sensor on each pin to receive signals from a video projector — is perhaps even more ingenious than the pins themselves.

[Leo]’s idea seemed destined for greatness, but alas, the cruel realities of scaling up struck hard. Each prototype pin had a low part count, but to be manufactured economically, the entire BOM would have to be reduced to almost nothing. That means an ASIC, but the time and expense involved in tooling up for that were too much to bear. [Leo] has nothing good to say about the patent game, either, which his business partners in this venture insisted on playing. There’s plenty of detail in the video, but he sums it up with a pithy proclamation: “Patents suck.”

Watching this video, it’s hard not to feel sorry for [Leo] for all the time he spent getting the tech right only to have no feasible way to get a return on that investment. It’s a sobering tale for those of us who fancy ourselves to be inventors, and a cautionary tale about the perils of participating in a patent system that clearly operates for the benefit of the corporations rather than the solo inventor. It’s not impossible to win at this game, as our own [Bob Baddeley] shows us, but it is easy to fail.

Thanks to [th_in_gs] for the tip.

43 thoughts on “Fail Of The Week: Bright Idea For LED Signs Goes Bad

  1. Wow that is an incredibly smart system.
    I think that would totally pay off but it would require to partner with big industry giants that already have established customers in the signage industry. Like he said, its a chicken egg problem so you gotta find someone who got chickens to start with.

    1. Also gotta find someone that doesn’t want to shitcan it so they can ride out the full 10 years or whatever of their projected product/tooling lifetime on their current stuff. Disruption is not welcome in many industries.

      1. Also need to find a big industry giant to partner with that’s not going to immediately decline the partnership and just steal the idea for themselves. You can put up NDAs all day long but if you don’t have the wallet to fight their army of on-staff lawyers you lose. They will walk so much process and discovery against you that as a small player you can’t win.

        1. well, some times it might cheaper to pay the inventor rather than the lawyers.
          That said, they most likely already have lawyers on their pay role, so they might as well flex a little.

    2. Soo I work with those guys with my company sometimes, they’re cheap mofos, I don’t see them partnering up for this. They’d rather just order LED strings from China, I’ve seen 20ft signage for major league baseball stadiums using cheap aliexpress lighting strings inside.

  2. I’ve wanted to make something like this for years, but like he says there: it is very hard to scale up. I’ve thought if it could be implanted on any surface, and programming it with the projector would give it information about where it is on the picture. As always, though power would be the issue, unless you could do it with some conductive substrate, or wirelessly.

    1. 1.) install off-the-shelf index-able LED strips and driver
      2.) install 1pc off-the-shelf Arduino ESP32 with sdcard
      3.) increment through entire address space once to flash each LED
      4.) Watch addressing process with a web cam in a dark room to resolve spacial mapping at each step
      5.) upload animation file to ESP32

      The software would take a few weeks, but the hardware is essentially open and generic.

      I don’t have time for another project, but OpenCV will make this easier.

        1. It is a tempting project, as for the scaling part I thought about a master sync broadcast UDP packet command setup in the file format. Thus, a local FreeRTOS mesh could sync up the animation frame’s start time to an arbitrary master unit, and the number of driver boards can grow rather large yet remain electrically isolated… Also, you still have the i/o bandwidth for each LED strip to do a decent frame rate in such a setup. So, the length of each run wold only be constrained by the LED strip’s baud rate, the 1/60s to setup the image pixels, and the sdcard data stream read out time. Notably, in such a setup a open collector or optically isolated sync drive input pulse wire between boards would also work with even lower latency.

          Maybe a video stream would be possible with a VLC plugin that used the pixel mapping file to broadcast over the limited Mesh bandwidth to queue up a few live frames worth of encoded LED strip data…

          I must focus on finishing my own stuff right now… as one must avoid getting distracted by other fun projects. ;-)

  3. Nice to see someone trying to come up with a high-quality option for LED signage. Most of it seems to be the cheapest-possible crap shoved behind what used to be Neon channel letters. The cheap stuff ends up with “LED rot” in just a few years.

    I love the outrageous LED stuff like used in Las Vegas, but to me, nothing beats good old fashioned Neon.

  4. This is a clever idea, especially the PSU based clock source. I would of thought with ever cheaper MCUs (especially not PICs) it would be viable. The programming method seems ripe for optimisation.

    1. Signalling with polarity changes is quite well know. For example OpenTherm does it. Return signalling is by modulating current consumption.

      Pixel mapping with a camera may be easier? Then once you have the map you can reprogram it as many times as you want with just a PC. This is also more common in the industry, the projector method seems novel.

  5. I agree with you. I watched the whole video and I went from “oh, that’s clever”, to “My god, this is way too complicated”. I think he completely missed the idea of minimum viable product to test the market, and sunk way too much time into features that would only be necessary after you’d determined that there was a strong market to support this type of thing.

    1. I stopped at 4:30 to say the pins only need to be slaves, all communication only needs to be one-way, the driver/PSU had* its own light sensor, it could have used even less flash memory to always already know how all its slaves had been programmed… it’s already another slave to the projector, just keep doing that! and all the syncing could’ve been done by injecting and sensing ripple on top of DC instead of having full wave rectification and smoothing *everywhere*.

      * a minute later I see there already is a sensor on the driver but I’m still saying nooooo open the loop

      ** I know there are lots of design decisions, scattered over time I didn’t spend, around facts of which I shall remain ignorant… and probably what I just said is only to demonstrate that ignorance… but my hard copy of Modern C arrived today, that means I’m automatically an expert in everything (/s)

      1. Also you can slash the internal complexity and the memory requirements, lower the power ratings on each pin– while you triple the demand for all but two components– if you split the RGB LED into smaller R, G, B versions and place a colored plastic window over a single photodiode. That could begin to alleviate the volume problem. Also they get smaller enough that maybe you don’t care if it looks like an old CRT’s triad shadow mask up close.

  6. Would it be simpler now to do it with say an ESP8266 chip. I mean the boards are ~$1 if you order high qty. probably could design it with LED and be under a $1. then do mesh network for communication. then use a simple webcam to look at design and create map of pixels. each one turns on an LED and the software maps out where it is. Power supply would be just a 3.3 or 5 VDC power, depending on needs. and the cheap 8266 comes with a 512KB, 1MB or 4MB memory right? 512KB would be enough to replicate this product. Speed up location by making each pixel a different color, a good 1080P webcam can pick them out. We all at home. who can build it first challenge?

  7. A clever idea to program the pixels indeed. Getting this to work solves massive technical challenges

    But what problem does this solve? What is the advantage?
    – I need a board and I need to drill holes. As others have mentioned – sticking addressable wired leds from the back is just as much work.
    – Calibration via projector or via camera – again no big savings for the end user.

    How to create an advantage?
    Maybe make the leds rearrangeable without a drill. Like with magnets. Think of a blinking restaurant menu that can easily be changed.
    I could imagine that with magnet leds would stick and can be rearranged. For power maybe one needs to use individual conductive feet.
    Imagine a module having three contact points, sticking to the board with a magnet. Can we arrange power supply islands that always two of them will provide power without short-circuit? Or would we need more and pogo pins?

    Today, I would imagine that you can omit the expensive ASIC route and go for a cheap microcontroller (starting from 1 cent) – maybe think about compressing the sequence.

  8. Hm, I don’t think this offers much beyond what you could already do with WS2812-type LED tape. You can stick that to a sign in any shape you want, and program it all from one end using an ordinary micro, no funky projector stuff. The only improvement this offers, is that you don’t need to have all your LEDs in one line. But this also comes with lots of disadvantages.

    The “advantage” isn’t even a big one anyway, you just choose not to illuminate any unwanted LEDs. So from a distance they may as well not be there.

    This idea is really solving a problem that barely exists, and not solving it well. The patent system might have done the bloke a favour, if it’s kept him wasting time or money on this thing. There’s a biiiig difference between an invention that works, and one that serves a purpose to the extent that people will actually buy it.

    Hm maybe our lad can compromise. WS2812s, in ordinary through-hole LED format, which I know they make. Then a big grid of holes they could connect into like a Lite Brite. With some sort of socket to put the data pins all in series, and connect through even if some sockets are unused, like headphone sockets do. In fact you could actually use headphone sockets! Do they make stereo ones with the extra contact on both? Because there’s your data-in and out lines!

    I dunno it’s a less stupid idea that a projector. Who’s going to bring all that rigmarole out, the projector plus the laptop and all the power supplies, to reprogram a sign that is supposedly reprogrammable but isn’t in any practical sense? What if you don’t own a projector? If you need to call the sign company every time you want to edit your sign you may as well just employ a signwriter! How’s that gonna show your latest bargains? Better than the LED grid signs that have been available for decades and recently got really cheap?

    Why not just use a fucking monitor!? With a computer behind it, or gods help us, a Raspberry Pi!?

    The real lesson here is some variant of “know what the fuck you’re doing”! Business is business. Things have to be feasible, not just work. And I bet “work” wouldn’t be 100% of the time with the utter Heath-Robinson affair of plugging the LEDs in then actually programming them. For a solution that would cost more than the monitor or the LED grid! Just…. no!

    Sorry dude, but this is a really terrible idea. Needs saying, I’m surprised nobody told the guy.

  9. I don’t see what advantage this product provides. You can get a 64×32 RGB matrix from China for 20$. Add an ESP32 for a nice web interface. For the same price I would only get like 20 of these very complicated high-tech LEDs and they are much harder to program than just uploading a new animation via a web interface.

  10. Unfortunately, I wished the video was more introspective. I have questions:

    1. Leo says patents suck, but doesn’t really say why. What was his specific experience leading to this conclusion?
    2. Leo says the endeavor failed due to mounting business complications and was a total waste of time, but what exactly were those complications? Were they avoidable given hindsight, or an indelible flaw in the product and business plan?
    3. What was the market analysis and business plan for this product, and why did they fail?
    4. While the concept is cool, was there a market need being addressed by this product? Are there sufficient customers who need an animated sign with sparse-but-fixed-position pixels that can be easily reprogrammed? Is the cost of a sign built using this technology competitive against, say, one with addressable WS2812-style RGB LEDs and a hardwired custom-programmed microcontroller?

    In the end I kept thinking: “that’s a neat idea but if I needed an animated sign, would I pay extra for this technology versus what was already out there?” Unfortunately, my conclusion is “no” because the sign is not a blank canvas–the overall shape and pixel placement are custom to the application and fixed and I wouldn’t need to reprogram the animation frequently if at all. How well would the custom projector programming system work if it’s 20 feet up on a pole or on the side of a building? I also doubt the cost savings of two wires vs. three would be sufficient to make manufacturing less expensive overall. If something more dynamic is needed, then perhaps LED video tiles are more appropriate.

    I do see a potential market for an addressable RGB LED that only needs two wires instead of three like the WS2812, similar to the Dallas Semiconductor 1-Wire system, without the program-via-projector feature. For personal projects, running less wires is good.

  11. This was very interesting to watch.

    I actually did a very similar project back around 1970! The project was to build a marquee sign for a Broadway show (starring Lucie Arnaz, as I recall). The sign was designed to wrap around the proscenium of a Broadway theater, and to show a variety of chase patterns between acts, In addition, the producers wanted to be able to display text on the marquee- I think the names of the cast members or something.

    I used a similar method- I created a scale model of the proscenium, with an array of photo-SCR’s that would match the full-scale array of incandescent signage lamps on the actual marquee. Then we would create masks between the array and a spotlight to create the patterns. Each SCR was wired to its corresponding light bulb on the large sign, and a house electrician would change the masks and turn on the spotlight to activate the sign.

    In order to create the chases, I proposed using polarized film squares glued onto plexiglass sheets for each animation, with a rotating polarizing wheel between the spotlight and the sheets (I had been building Kodak Carousel slides for a while using this technique- it seemed scaleable). Finally, I decided to use fiber-optic light pipes between the proscenium model and the photo-sensors to increase the density of the model.

    I remember driving hurriedly from Broadway in NY City down to Edmund Scientific Company in Barrington, NJ to pick up the polarizing film and fiber optic cable. I got there after they had closed, but Dr. Norman Edmund himself came out to give me the materials, because he was fascinated by the project.

    Initially, it was a disaster! First, I discovered that the SCR’s were only rated for 100 Volts! We fixed that by putting the entire marquee on a big dimmer! Then, the sensitivity was pretty poor- especially due to the fiber light losses, so we needed a 1000 watt spotlight to compensate. Finally, there were issues with RF interference between the SCR’s, which threatened to undermine the entire project. Somehow, we finally got it working in time for the show to open!

      1. Was that the same ES as the catalog of the 1980’s? My dad was a science teacher and used to give me his Edmond Scientific catalog when I was a kid. I used to order “bags” of DC motors and solar cells to play with. Was always intrigued by their lasers and fiber optics. Used to imagine all the things I could do with them if I had the loot.

  12. I did not watch the whole video, as from what I gathered this system is far to expensive because each pixel needs its own uC with enough memory to store a complete animation (for that single pixel). Also changing animations is a bitch. Also, The aluminum sheets with plastic foam as power source have a dis-similar metal interface with the LED cups and screws. making weather resistance needed for outdoor sings unreliable.

    Why not simply turn it around.
    The LED strings he shows at 00:58 are cheap from ali, and he tries to play them down as difficult to map and quickly goes on.
    However, these are fully encapsulated, come in water resistant outdoor varieties, and are also very easy to map.
    Just put them in random order in whatever object you fancy, connect them to some uC (Those leds are something like WS2812 compatible) write a simple program to light them up one at a time, and point a cheap video camera at the sign.
    Then extract pixel placement and order from the video data.

    Then you can make any animation you want, have live updates, etc, and you only need a single uC.
    Much more convenient and cheaper.

  13. Looks to me like he went down a rabbit hole too quickly and got stuck in there. Sure is interesting to not just have a concept in mind but actually follow it thru and get something working on the bench. But somewhere along his path, he just lost the view of the bigger picture and did not revisit and reevaluate his initial design choices anymore…
    If you already have to make a PCB with a custom ASIC chip and an RGB LED, the next logical step would just be to have a fully integrated microcontroller in the LED-package. So you’re just making a more advanced “WS2812” style LED.
    If that would have been a viable option, i’m pretty sure the big serial LED manufacturers would have picked up on that idea long ago.
    To solve the data connection to each LED, i’d probably try to use a DC voltage supply (24V maybe) with a voltage modulated datastream (5Vp-p maybe) on top. If each LED had a preprogrammed unique number, you would just have to map out the unique numbers once (could probably be done when you assemble the panel and place each LED) and feed that to your display animation software.

  14. What’s the market? A minimal video-resolution system will have a retail price at least in the $75K ballpark. There’s a big markup between parts cost and assembled, tested units with amortized costs, overhead and margin for all the people who touch the system. Even a small system will be expensive. Don’t fall in love with the technology. Be pragmatic about the final prospects from the moment you consider the idea. I have a hundred ideas on my list right now. Viable ideas? Maybe two. Ideas that will make it to market? Maybe 10% chance of one successful product. Depends on your definition of success. It never ceases to amaze me about the idiotic stuff which becomes successful and the good ideas which fail. If I could have one really stupid successful idea I’d be happy. I always wind up with the other kind.

Leave a Reply to lionxl Cancel reply

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