New Part Day: Two Millimeter Addressable LEDs

The WS2812, or “Neopixels”, or whatever you want to call them, are the standard when it comes to adding blinky to anything. These chips are individually addressable RGB LEDs, which you’ve seen in many LED strips and a thousand other products. These LEDs are rather big compared to normal, dumb LEDs, measuring 5 mm on each side. Here are WS2812s packed into a 2 mm x 2 mm square package. It’s the smallest and brightest blinky that works the same as the WS2812s you know and love.

This is the latest product from Worldsemi. We’ve heard of these before, but damned if we could find a supplier or even a price. Now they’re on AliExpress, at a price of $8 USD per 100, shipping not included.

Electrically, these appear to have the same properties of the normal, 5050-size WS2812 LEDs. Apply power and ground to two pins, send data in on one pin, and connect the next LED in the strand to the remaining pin. Yes, it requires a bit of work to turn this into a display, but microcontrollers are very fast now and have plenty of RAM. Attach a BeagleBone and you’ll be able to drive as many as your glowing heart desires.

If you’re wondering what the coolest project imaginable for these LEDs is, here’s the math: the largest (common) PCB panel for your random board house is 16 by 22 inches. Assuming a 3 mm pixel pitch, that means the largest PCB display you can make with these LEDs is 135 by 186 pixels, call it 120 by 180 just to make things easy. That’s 21,600 LEDs, at a cost of about $2,000. I would not recommend reflowing these, and assuming soldering a LED every thirty seconds, it will take about a month to solder them all by hand. There’s your project, now get to it.

52 thoughts on “New Part Day: Two Millimeter Addressable LEDs

  1. This is nice. But real “missing link” currently is 24V addressable leds. Those 5V leds draw so much current that it is practically impossible to make anything large with them. I dont know why they do not make leds where is smaller leds as series inside of one packet and controller.

      1. No. An LDO burns the voltage difference as heat. Your LEDs efficiency degrades down to the level of the ancient incandescent bulbs.You do not want heat with LEDs. That would be the worst and most useless solution.
        Construct a device with point of load 24V to 5V switch mode converters, if you need 24V operation.

    1. I agree – 12v strips are addressable in 3’s so I don’t get the resolution needed. Would be nice to have cheap individually addressable 12/24 volt LED chips. My 12 volt supplies are half the size and give more power than 5 volt ones. I don’t get quite as good returns on size/cost when going from 12 to 24 so 12 is the sweet spot.

    2. The packaging will get very complicated. In the current design, all three LEDs share a common terminal, they are placed on the same metal part which is VDD. If you want 24V power, you will need to put 6 of each color minimum in series requiring a lot more metal frames, bond wires and space. It would be too expensive compared to thickening your supply rail.

      I do however agree with you that most strips with WS LEDs have poorly sized power rails, but that is what you can fit on a strip.

    3. These are still fundamentally LEDs, at 24V they don’t shine they burn. If you want to drive multiple LEDs with 24V you’d be able to use the WS2811 controller but you’d have to drive them as one, and you’d need MOSFETs and something to drop the supply voltage for the controller. It doesn’t seem worthwhile.

    4. Shiji Lighting used to make 5050 size addressable LEDs that were 12V and had 3 LED dies per color in one package. I managed to get 1m samples for the SJ1211 (single data in/out) and the SJ1221 (dual data in/out) before they quit production. A real shame. These things were BRIGHT at 20mA per color.

    5. I have some “Christmas” addressable led strings made by an italian company named LedWorks, and sold under “Twinkly” brand name in some european countries (i’m in France).
      https://www.twinkly.com/fr/

      Maybe also under “LeaveUp Lites” brand in the US (unsure, but found an app for them developed by the same company in the iOS appstore).
      https://leaveuplites.com/

      These strings use only 2 wires (so power and data are combined), and a 24V power supply.

      I’ve not yet took time to have a look at the output signal with an oscilloscope, but since i wonder how i could drive them by myself, i will… maybe… one day!

      The controller is based on a ESP-WROOM-02 module.

      I don’t know what are the LEDs used in them.

      As a lot of IoT devices, they are quite insecure:
      https://labs.mwrinfosecurity.com/blog/twinkly-twinkly-little-star/

      But they work quite well (despite software could probably be improved), and they can be associated in groups to spread effects accross several of them.

      1. Why? Stencil print 21000 pads and PnP 21000 LEDs.. One pass in an eazy-bake oven and done! Use the rest of the 29.6 days of the month to enjoy LED brilliance ( @ 6.3 peak KWatts!). There is even a reflow profile in the Aliexpress order page.

    1. I think he meant to say “I would not recommend handsoldering these” and the rest of his article is why you shouldn’t…. Reflow is the only proper way to mount these.

    2. These LED’s are notoriously fickle when reflowing. I lost an entire initial test run as the contract house didn’t follow the reflow instructions EXACTLY. Even when they did, we still ran into issues and had to hand solder replacements in. I know Adafruit did their neopixel products by hand for this same reason. Not sure if they still do.

      1. I reviewed the reflow specifications on the Aliexpress page and there is nothing there that isn’t reflow soldering 101. I am really curious as to what the issues are with these. My experience is with higher wattage RGB LEDs from Osram, Edison, etc (1-3 watts) and although they are similar in design, reflow issues were nonexistent.

      2. It could be that they don’t meet the profile in the datasheet for various reasons – knock offs, rejects etc
        Someone in China probably figure out the right profiles as they are happy selling those things in reels.

        1. In my experience the biggest issue with these for reflow is their MSL (moisture sensitivity level)… You _MUST_ keep them in a dry cabinet. They have a 24 hour shelf life in ambient air and I would _NOT_ want to push that boundary. These things seem to suck in moisture to all the worst places in the chip and make a depressing snap crackle pop sound as you hit them with a hot air gun if they’ve been out too long. (I opened a package of 10 from Ada, got through 8 of them day 1 and decided to go ahead and try my luck with the remaining 2 on day 2… Yeah, for those of you that remember an old TV show called “Hee Haw”… If it weren’t for bad luck, I’d have no luck at all. They went pop almost instantly upon applying heat to them. The pop sound happened several seconds before the solder even started to think about melting.

          They are listed as MSL category 5A (at least the APA102 series). I haven’t looked at the 2812s or similar ilk, but I can’t imagine why they’d be different. LEDs in general tend to have higher MSL than other similar sized devices.

          So I agree that reflow is the way to go with these, especially for a large panel with a lot of them, but you’d have to do it with a pick and place… No way you could get all those LEDs hand placed within the MSL limits.

          I confess, I’m half tempted to buy a reel of 2020 APA102s just to watch the P&P hammer them out.

          I did buy 100 of them on cut tape from AdaFruit. Problem is that the cut tape is too small for the machine and hand placing them wasn’t within my limits of dexterity.

      3. I also had issues with reflow using standard solder paste, even with prebaking. My guess is that the internal connections break.

        If you use a low-temp solder paste – I use tin/bismuth/silver – they are easy to reflow.

  2. Hard pass on this no clock line nonsense. Also 2020 smart LEDs have been out for a long time, I do have plenty of APA102 2020 LEDs. What these products could really use though is support for Vdd of 12V or more (they already have a built-in regulator so I don’t know what the holdup is) and a Vio guaranteed to to work down to 3V.

    1. I’d rather have them work down to 3v than with 12+. I have more occasions where I’d like to drive one or a few addressable LEDs from a single lipo cell than I do needing to drive large numbers of them. Hate having to provide 5v just for the LEDs, or run everything at 5 and then translate i2c down for 3.3v only components.

      1. I know this isn’t quite the same, but I have a project right now using the SK9822 (the APA201C copy) working at 3.3v. Just putting that out there in case it’s a viable option for you.

    2. I know the specs say they require a VIO of 5V, but I’ve never encountered a problem driving the APA102s with 5V power and 3.3V Clock/Data. Of course, YMMV and it is out of spec, but it’s proven reliable across a lot of LEDs so far in my experience.

  3. When is some company going to make something like a 64×64 or 128×128 RGB LED pixel panel with seamless edge to edge butting? Fabricated all in one piece it would make it possible to build VGA and higher resolution displays without having to be as rich as Zuckerberg, Gates, or Trump.

    307200 pixels for 640×480 VGA. $24576 at $8 per 100. Surely there would be some discount for ordering 308000 of these LEDs?

    1. You are making a very bad design decision. When you have an array that big, you should at least try to use the much cheaper dumb RGB and design the driver circuits for the whole array. These “toys” aren’t scalable in cost/update rate.

      You know something like row/column drivers like those charleplexing but much bigger. This is a case of high I/O count FPGA comes in handy. There are probably industrial suppliers that make these things.

  4. I was looking at the feasibility of creating a very small buck converter that would fit under (so the LED package would piggyback on top) the 5mm x 5mm version of the LEDs. Never did the efficiency analysis to see if it made sense, but it makes me wonder. Then you could potentially feed in a range of voltages and not have to worry as much about voltage drop.

  5. Personally, I like the APA102 (specifically APA102C) products quite a bit better than the WS2182. These are available in both 2020(2mm^2) and 5050(5mm^2) packaging.

    The huge advantage to the APA102 product family is that they use 2 pins (clock and data) so there’s no sensitive timing and a brief lag in clocking out a bit or two due to servicing an interrupt or processing wifi data or whatever is a non-issue (this seriously improves reliability on systems like the R.PI and ESP32).

    It also allows for much faster clocking of your LEDs… Upwards of 24Mhz and they have a much faster PWM clock than the 2812 and 6182 products I’ve seen (yielding less perceptible flicker, especially at low brightness values).

    Adafruit sells these under the “DotStar” name. If you want whole reels of them, Shiji Lighting in Shenzhen is where Adafruit gets them.

    One potential solution to the voltage problem is to inject multiple 5v sources along your string. You can either do this with 5v segments, or, you can diode-isolate your regulators from being back-fed from the power bus. As long as the grounds are all tied to the same level (one common ground circuit) either solution works. Segments will make for slightly more efficient operation, but diode isolation makes for greater configuration flexibility.

    You can power about 24 of these from a 2A 5V source (or ~5.4V source if you’re diode-isolating with a low drop Schottky).

Leave a Reply

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