#246

[@bardenegri21]

I never had a ws2812 strip that turns on full bright without a signal. It must be some different chip or a new revision or something. I have a big ambient light based on neopixels in my living room, so here are some more tips for long wires. Put a 1000uF capacitor across the power pins every two meters of a strip to reduce chance of flickering. 3.3v signal levels only work for very short wires, for longer you need a level shifter. For more than 2 meters you need a gooood level shifter and coax. I used an octal bus transceiver chip like adafruit recommends and still needed coax. A reverse diode and pullup works well for shorter distance. For high current you can run a thick power wire to both sides of a long strip for best results. Fastled's current limiting is great as an added protection. You can get neopixels in a classic through hole led package now (5mm or 8mm), they are amazing! I'm making Christmas lights from some, asides from the extra wire they look exactly like normal lights just way brighter.

[@Ruudrad]

I sofar built 5 wordclocks using LED strips, each with 5m WS2812, 33 LEDs/m. The first strip indeed had this "on" characteristic: All LEDS white and on. I thought the others were "defective" as they did not switch on white, but they worked as specified reacting to the data-in, using the same library, python program (on RPi) and parameters. Later strips were for sure WS2812B, do not know for sure for the first one. Also the first strip was capable of showing distinct colors for the lower intensities slightly better. I guess there is other hardware involved.

[@crayzeape2230]

The length limitation at 3.3v is more likely caused by vcc dropping too low over distance, and not because the data signal is too low. The data line from the MCU only ever drives the first WS2812, which in turn drives the data line of the next, and so on. The "data out" voltage is generated from the vcc connected to the WS2812 and is not related to the "data in" voltage. You can test this by feeding 3.3v signal levels to the "data in" of a 5v powered strip and checking the data line voltage further up the strip, you'll find 5v signalling.

[@bardenegri21]

Exactly. I had a bit of a distance between the controller and the strip, hence the problems. You can also add a blind pixel next to the controller to act as sort of a level shifter but even then the drive current isn't good enough for really long wires. It's also the capacitance of the wire, it acts like a filter for the fast signal rounding all the peaks into oblivion.

[@crayzeape2230]

@@bardenegri21 Good point, I was thinking only of the controller being electrically close to the strip. I wasn't considering that there may be a long run of cable between controller and strip.

[@wkharmon]

I’m using what appear to be the same matrixes that he is using and I have not observed the “on” characteristic that he is seeing either. The pixels remain off until they get a data signal. It’s interesting that similar hardware would behave differently. I am using the cap and resistor that are recommended for NeoPixels, as well as a logic level shifter to get it to 5v. I was curious if that could be causing the different behavior, but it sounds unlikely.

[@plemli]

Tip #14: find neopixels based on the WS6812. These are easier to drive, and exist in RGBW (cold or warm white) when you want more purer white or nicer pastel colors. The disadvantage is that when only blue is on also makes the phosfor in the white part glow a bit, so the blue will be less pure.

[@AndreasSpiess]

Thank you for the tip! In my case I had to go with the panels I got :-(

[@leachim66]

"my old counter only had 5 digits" how humble of you Andreas, congrats on 100k! (+)

[@AndreasSpiess]

I only had a few subscribers when I built it .

[@plemli]

Tip #13: add a small and cheap 24 to 5V buck converter (less than 1$ for a 2A one) to each set of n neopixels, and use a 19v laptop power supply to feed the lot. I did this to avoid voltage drop and the ability to use much thinner wires.

[@Ruudrad]

What I did and confirm to be working. Also use a _separate_ buck convertor for your RPi or Arduino.

[@cabe_bedlam]

Having recently designed an LED system to go into an SUV for a ridesharing company, can confirm this is the way to go. "Why do you need 30 PSU's and two batteries?" "Because otherwise, the cables will be as thick as a hot dog....."

[@georgekot6377]

plemli -> Great idea, thanks !

[@danheidel]

I would add a word of caution on using the 2812 chipset. I've done some professional work a couple years back on large installation LED pieces and the 2812 LED strips were neverending nightmares of unreliability. You tend to get dead pixels after a few weeks of continuous operation that kills all signal transfer downstream of that point. This resulted in some catastrophically expensive callbacks. We tried multiple suppliers and saw the same issue come up repeatedly to varying levels. And yes, we followed strict ESD protection protocols as well as using custom boards with built in overvoltage protection diodes. We ended up having to fall back to the older 8806 chipset. Those chips require a data and clock line, but have proven to be far, far more reliable in the field. I don't know if the 2812s have improved at all in the last couple years but I would recommend sticking with the 8806 chipset if the project requires long-term reliability. If you do use the 2812, I would recommend buying the LEDs early and running them 24/7 with a continuous pattern for at least a month to catch dead pixels so you can cut them out before installing them. In my experience that tends to catch most of the pixel failures but some will sneak through even with a month-long burn in.

[@zephsmith3499]

Ah, thanks for that tip. I suspected there must be some common failure like that behind the development of the WS2813 and similar. These versions have two data in lines - one is connected to the data out of the previous chip, and the other is connected to the data out of the pixel before that. If a chip stops working, then the following chip will get it's input signal from the chip previous to the dead chip, and the rest of the display will work. So it's intended for exactly the situation you describe.

[@danheidel]

​@@zephsmith3499 Ah, interesting cludge to get around the dead pixel issue. The problem is that double bad pixels were not that uncommon an occurrence in my experience. I had no way to analyze the failures, but if I had to guess, it was substandard ESD protection on the 2812 chips. Strips that come in the silicone waterproof jackets tended to have considerably lower failure rates as long as the jacket was kept unbroken. As soon as you handled an emitter, the probability of long-term failure went up. That was very noticable when we cut out the dead pixels and the failure rate in the neighboring pixels was around 30-50% in the next week. I'm guessing the 8806 chips used a older, larger fab process and can take static electricity hits better. If one has to use the 281x chipset, I would definitely recommend going with the 2813 chips with the ability to route around dead pixels, getting the kind encased in silicone waterproof tubing and making custom driver PCBs with zener overvoltage protection diodes on the data and power lines. Reverse voltage protection diodes would also probably be a good idea too. Also, I would still recommend the month-long burn in as even strips with the most delicate handling still had occasional failures. Don't cut the strips or breach the silicone tubing unless it's absolutely necessary in your design. And be sure to order a few extra strips so they can be hot-swapped in case of failures. If I ever do LED strip work again, unless something has drastically changed, I'll stick with those old 8806 chipsets. They're not as capable and need double the data lines but at least they have a reasonable reliability.

[@AndreasSpiess]

My application is not life-threatening, but it runs quite some time every day. I will report back if one of my pixels dies.

[@guatagel2454]

Andreas, every time I search for some technology that is new to me, your channel shows up. Thank you!

[@AndreasSpiess]

You are welcome! So we are both curious people.

[@kevingovender2656]

We wont stop watching "the guy with the swiss accent", keep up the good work, however everytime you say "dirt cheap" the Chinese push up the price 😱

[@AndreasSpiess]

So far they had enough internal competition ;-)

[@Ruudrad]

Remember, that in the world of the Swiss, even if the Chinese up the price weekly, they will stay "dirt cheap" for quite a long time ;-)

[@casio007]

but sometimes it needs speed up the video xD

[@kevingovender2656]

@@Ruudrad Very informative in the " Swiss world" everything is cheap, there will be no need for me to join this channels Patreon page

[@Ruudrad]

@@kevingovender2656 In case you didn't notice, it was a joke and not meant to be an indicator for contributing, which I feel should be related to what you think of the channel quality. Which by the way I find rather good.

[@Gunther0867]

Hey, Schweizer! Danke für deine Videos! Kompetent, relevant und humorvoll! Super! Weiter so! Baai

[@AndreasSpiess]

Vielen Dank!

[@mattytrentini]

As for controlling Neopixels with MicroPython, it is already supported on the ESP8266 and ESP32. Both are implemented with a bit-banging solution. For the ESP32 I intend to create an alternative implementation that uses the RMT module so that timing is *very* accurate. The Loboris fork of MicroPython already has an implementation that uses RMT so Boris has already paved the way; I just want to implement a slightly different design. It's also worth noting that FastLED has an RMT implementation for ESP32. Using RMT should be a significantly more accurate way to drive Neopixels than any bit-banging solution - accurate timing becomes important for longer (1000's) of Neopixels. If anyone is interested in this please reach out!

[@waheedidrees]

15:27 some emotional stuff.. Long live Man with Swiss accent...

[@AndreasSpiess]

Thank you!

[@davidstonier-gibson5852]

Viel Dank, Andreas. I appreciate your nice low key style and lack of duffa duffa music. Just good information provided in an easily digested form. I have only recently discovered these pixel LEDs and I am blown away. The first LEDs I used in my work cost $1.05 in 1969 dollars, and you almost had to turn the room lights off to see the dull red glow. have also designed architectural LED effects lighting systems, but it was hard work and involved a lot of processors. NeoPixels make it so easy!

[@AndreasSpiess]

I completely agree: These Neopixels are the best invention after "geschnitten Brot" ;-)

[@davidstonier-gibson5852]

@@AndreasSpiess We say "best thing since sliced cheese". :-)

[@zvpunry1971]

That beeper makes it somewhat tempting to write a script that non-stop unsubscribes and subscribes. ;)

[@AndreasSpiess]

Pay attention: Google uses AI ;-) They just wrote me that last week they cleaned the platform and removed accounts which did not behave as they like... In my case, they found more than 100.

[@zvpunry1971]

Platforms like KZbin have to remove accounts all the time, one example are spam-bots that comment their spam on all top-comments. By the way, I like the German abbreviation "K.I." more, it allows me to say "Keine Intelligenz". ;)

[@AkosLukacs42]

I have an idea, just unsubscribe / subscribe when someone comments here, and add a comment yourself, so it will be a nice chain reaction :)

[@oskimac]

That comment was (useful or interesting)

[@Valenorious]

@@zvpunry1971 Well, "absent intelligence" works also. ;)

[@e74av]

Comprehensive as usual... Even with just a few new info, worth watching every second! Simply because you try not to miss anything important for a particular topic! THANK YOU

[@AndreasSpiess]

You are welcome! And Merry Christmas!

[@mr.b.1589]

Connect the first +5V/GND (in) and the last +5V/GND(out) the your power supply, to fix the problem with different brightnesses. AND you don't have the full current on one cable pair.

[@AndreasSpiess]

Like that you still would have the middle matrix with less voltage (longest distance from the power). If you connect GND on one side and VCC on the other then all LEDs have the same distance from power (not my idea. A viewer wrote it).

[@mr.b.1589]

@@AndreasSpiess That's interesting. I'm using an LED strip with the sk6812 with about 100 LEDs as I described it. I didn't notice any difference in brightness. Maybe it occurs when I use more LEDs. The advantage of the strip is that you can connect several other wire pairs in the middle.

[@cannesahs]

Lazy makers? ;) Run vcc to gnd wires to each matrix

[@AJB2K3]

Lady Ada advises to put a capacitor on the supply lines because of that initial boost issue.

[@AndreasSpiess]

Without a FET, in my opinion, a capacitor increases the surge current, because a discharged cap behaves like a short-circuit in the first moment. Only with a FET you can charge the cap first and offload the PS. Maybe she wants to help with varying numbers on pixels ON during operation. There, a cap without a FET helps.

[@arduinosynod1571]

Rather to put ntc to prevent surge current. Andreas got point.

[@opsahle2]

Andreas Spiess The cap could have a series resistor towards the power supply to reduce surge current.

[@CatLighting]

The controller can be powered via a diode and insert a capacitor behind it.

[@joeybushagour2612]

@@AndreasSpiess they mean a capacitor across the ground and vcc line. This will help "absorb" the initial surge of current because the capacitor can supply the rush current.

[@roelskiunplugged1134]

As allways: super informative and professional! I love your content!

[@AndreasSpiess]

Thank you!

[@santorcuato]

Thank you Andreas for your sense of humor and your way of explaining things. I hope you get to use the next digit ... And that I see it!

[@AndreasSpiess]

Thank you very much for your nice words!

[@electron-1979]

The NeoMatrix library rocks! (Please support Adafruit by buying their products, at least sometimes).

[@RichardT2112]

We love the guy with the Swiss accent! Merry Christmas Andreas!

[@AndreasSpiess]

Merry Christmas, too!

[@lluisllimargas2637]

Teensy MCU are an excellent option for projects related with Neopixels & WS2812 LED strips due to its fast Direct Memory Access and extremely precise timing too. Thanks Mr. Spiess!

[@AndreasSpiess]

So far I never used a Teensy. Thank you for the tip.

[@SingenDrake81]

What are the ways you use to attach the metal bar to the pcb ? Is it just thermal adhesive ? Do the ws2812's run hot at all (prolonged white light) ?

[@mattytrentini]

Continuum if they are driven at full white each pixel can draw 60mA so yes, they run hot! Power consumption does drop off dramatically when the brightness is reduced or less colour channels are employed so it’s typically less problematic in the real world.

[@HiVisl]

You are a great teacher. Very knowledgeable. I really enjoy your channel, sense-of-humour and amazing knowledge. Thank you for all you do.

[@AndreasSpiess]

You are welcome. And thank you for your nice words!

[@MakunaRGBIC]

NOTE: The Adafruit library will have issues on Esp8266 when there is WiFi traffic as it is BitBang and the WiFi interrupts will "interrupt" it (which can not be disabled without crashes). I know, as I wrote the original and created NeoPixelBus to fix it.

[@AndreasSpiess]

So far I did not encounter problems. It runs now for a few weeks. Maybe because I only contact KZbin once every minute or so and only update the pixels afterward. But it is good to know if I run into problems!

[@Webshadow]

"Gosh dangit", or whatever you are supposed to say, that was very useful for my 3D-Printed LED-Cube-Project, i will however stay with my ESP32 to get BLE for connecting to an app on the phone

[@AndreasSpiess]

I think there are libraries which work for the ESP32. But I needed one which works with the matrix library.

[@GRBtutorials]

That font has one problem: there's an empty row at the bottom, but not at the top. Ideally, there would be empty rows at the top and the bottom, so as to make it more beautiful.

[@AndreasSpiess]

That can easily be done by changing the cursor. I will try and compare. Thanks!

[@John_Ridley]

The empty row at the bottom is to allow for descenders. If you don't allow for that, things start looking really horrible when you have a character with a descender since it has to be shoved up above the baseline to make room for it. We had this back in the day when the first 5x7 matrix printers tried to print lowercase.

[@AndreasSpiess]

For the moment I only have numbers. So it should be ok.

[@psubond]

Awesome video. Only suggestion is use the mario brothers 1 up sound when you get a new subscriber and a mute button

[@AndreasSpiess]

Good idea!

[@AntonioBdeJesus]

Andreas Spiess strikes again! Thanks Guy! I'll experiment this, for sure. I did not know Neopixels yet.

[@AndreasSpiess]

They are worth the money!

[@confusedwolf7157]

1st time! RTFM (had to do double take)!!

[@ollieb9875]

Was it not the FAQ (fac) and the manual. 🙃☺️🤣

[@AndreasSpiess]

:-)

[@sethrd999]

Ive been looking for something like this to brighten the office, cracking project and well presented with your findings / fixes. Using the information in this video I dont see there being any issues that will stall it out any..

[@AndreasSpiess]

Such displays are really quite colorful :-)

[@springwoodcottage4248]

Thought about using neopixels several times, but The Scrouge in me struggles with using all that current even if it is small compared to what we used to use in much cruder displays, but the effects you show are so nice that I probably should use them. Thanks for sharing!

[@AndreasSpiess]

As I said: I love them because the excitement per effort is rather high :-)

[@plemli]

Tip #15: if your power supply cannot provide enough power (say max P1), calculate the estimated power draw P2 before updating the pixels, and scale the rgb(w) by P1/P2 so the max power is never exceeded even when the sketch would try to send a too intense image.

[@AndreasSpiess]

Good idea if you want to save on the power supply (and your LEDs do not go on at startup)

[@OsoPolarClone]

The color of the neopixels in your display are white which means all 3 LED’s are on. Will the current be less (1/3) if only 1 color LED is turned on?

[@AndreasSpiess]

I did not measure, but if I remember right I dis not see big current fluctuations. So I assume the chip adjusts the max. Power. Maybe somebody knows more?

[@mattytrentini]

Last time I measured it the current is indeed less for single colour channels.

[@zephsmith3499]

Yes, given a power supply that can handle full white, it will draw about 3 times the current (and power) of full red, green or blue. (However, as you can see in the video, if the supply voltage sags, the curreent decreases, so in that case full white would draw less than 3 times the current of full single-color). The other caveat is that the current draw is non-zero for full off; that is, each pixel tends to draw something like 1mA when unlit, to run it's digital logic and internal voltage regulator. I call that the "dark current", acknowledging that the term has other uses in other contexts. (And some pixels types draw more than that). So total current consists of a smallish "dark current" base plus a variable amount which depends on the brightness of each color.

[@PvGeens]

Tip # N+1: Powering a long series of Neopixels with even power distribution: Due to the resistance in the (thin) power lines, the pixels at the end may be a bit dim. The pixels at the end suffer from double resistance in BOTH power and ground lines. Suppose 1 Volt is lost in a single line, then the ground level at the end is 'lifted' 1 Volt, and the last pixel gets only 3 Volt. When connecting VCC at the END of the line, each pixel has exactly the same total wire length (and resistance) to the power unit. So each pixel gets the same voltage, regardless its place. The first data line level is correct compared to ground level. At the end, the ground level is shifted up 1 Volt, but also is the DataOut level of the previous pixel. So far so good. Then add an extra 7805 at the end of the series before feeding it to VCC of the strip. Connect the ground of 7805 to the (shifted +1 Volt) ground level at the end of the pixels. This creates a +6V VCC at the end of the line, resulting in exactly 5V compared to gnd (locally), regardless the position in the line. Even at the beginning of the line. (the 7805 must be fed 8 Volts or more). Temporal extra (or less) current drawn by bright (or dark) LEDs may result in more (or less) Voltage loss in the line resistance. This will be compensated immediately by the 7805. You can still add capacitors on the 7805 to reduce oscillation, and some current rush in limitation as described by others. You still need thick enough wire to prevent fire, but you don't need to worry about dimming one side anymore. The same trick (feeding from 2 sides for even power distribution) also applies for 12Volt garden lights (otherwise the last one gets almost dark), or speakers in a sound system of a large building (the last one produces almost no sound). Good luck.

[@AndreasSpiess]

Thank you for your thorough explanation. The only disadvantage is that you potentially need more wiring (to both sides instead only to the middle point). But for bigger installations this is worth the effort.

[@XTronical]

Great vid on these little beasties, I've never looked into them yet but this will help me a lot to get going.

[@AndreasSpiess]

:-)

[@BerndSchmitt-Martinique]

Unterhaltsam, Informativ und umfassend. Sie müssen einen exzellenten Drehbuchschreiber haben. Liebe Grüße aus der Pfalz

[@AndreasSpiess]

Vielen Dank für die netten Worte. Wenn das meine Sprachlehrerin sehen könnte dass ich noch Komplimente als Drehbuchautor bekomme. Sie ist ab und zu fast verzweifelt...

[@alaskawoodman]

Andreas, your videos are great. Thank you for your efforts and knowledge.

[@AndreasSpiess]

You are welcome!

[@gaborungvari784]

RTFM:) that was good:) thanks for the video and the resources!

[@AndreasSpiess]

You are welcome!

[@lloydrmc]

Interesting how far this technology has come in the intervening years. I've been watching another channel where he builds outdoor light displays, sometimes by hacking commercial RGB light bulbs/fixtures. Meanwhile, Andreas is at 400k subscribers. I wonder if he still uses this counter?

[@AndreasSpiess]

KZbin changed their policy and only change the counter every 1000 subscriber. So a counter is no more interesting :-(

[@electronic7979]

Nice video 👍

[@AndreasSpiess]

Thank you!

[@GhVost]

Very nice video as usual, Andreas! I believe you should get all seven PCBs. Good luck with your channel!

[@AndreasSpiess]

Thank you!

[@pedromms8908]

Very nice and simple display but you forget one thing: just add a mate acrylic sheet in the front of the LEDs. With that it will be much easier to see the graphics!

[@AndreasSpiess]

Good idea. I will try it.

[@donpalmera]

For the connectors you can use the same pin headers but push the plastic down until all of the metal is on one side of it. Then you can push it through from the front, solder on the back and still have the length required for mating connectors.

[@AndreasSpiess]

You are right. This is a good solution if you do not have a small tip

[@KirinDave]

A year later I am still left wondering how you actually did the wiring and affixing of those panels to the aluminum brackets.

[@AndreasSpiess]

The wiring should be clear. Otherwise, you find many other sources that show it in maybe a more understandable way. And I used double-sided sticky tape to fix them to the aluminum.

[@LucasHartmann]

A thick power supply wire connected to all panels should fix the voltage drop. You can do the same on long strips, connect the supply to several locations.

[@AndreasSpiess]

You are right. In my case it was not a big issue because most pixels are off and the power consumption was only 0.6 A

[@norm1124]

100k ? - Bald sind es aber 300k :-) - Gratulation aus Bern

[@AndreasSpiess]

Danke! Ja, es sind bald 300k. Ein Forecast hat mit 1 mio auf meinen 66 Geburtstag versprochen. Mal sehen...

[@sblop]

Great video. I am working with neopixels at the moment and this is a great video.

[@AndreasSpiess]

Thank you!

[@altamiradorable]

Danke fur alles Andreas ! Love the swiss accent ! ;) Nice video ! Exactly what I was looking for !

[@AndreasSpiess]

You are welcome!

[@robatoto]

NeoPixels that are fully on at power up seem like a real pain. I wonder why the purpose is. Luckily I never encountered this kind in the many projects I did. It would have been quite an issue as I often did‘t have sufficient power to run them all at full power. Instead I used software to limit the combined brightness in order not to exceed the max current.

[@AndreasSpiess]

You are right. I also think this should not happen.

[@TheTruthSentMe]

It's probably so you can use them as "dumb" LED panels without the need for a controller.

[@electron-1979]

A p-ch FET will be on during boot on a RPi. It boots with GPIOs inputs with pull-downs.

[@CreativeJE]

you will grow man we need more videos to learn and u are doing a great job thanks for these awesome videos which we dont find on other channels :D

[@AndreasSpiess]

Thank you!

[@LautmannStinkebart]

Sehr toller Channel! Grüße aus dem Neckartal

[@AndreasSpiess]

Danke!

[@MatejBosansky]

I get very good results with wemos D1 mini(esp8266) and fastled library. When i compared fastled and adafruit library, fastled changed colors much faster. Fastled has also many advanced features. For your project is suitable also micropython. Another good practice is to use max brightness of 1/2 value. You can save a lot of current and difference between 1/2 and full brightness is not so markable.

[@AndreasSpiess]

Thank you for the info. Unfortunately, I had to use the Adafruit lib because I also wanted to use the Matrix lib.

[@dan-nutu]

Congratulations, your "R number" is clearly greater than one if you went from ~104k to 224k subscribers in 18 months. Very well deserved! If the "contagion" keeps spreading with the same pace, how long before you'll need a bigger display?

[@AndreasSpiess]

Of course KZbinrs always hope for a "higher pace" ;-)

[@elektrofan5918]

Hello, I really like how you explain and that's why I come up with the idea that a video with a Matrix (ws2812b) JINX GLEDIATOR (PC) + ARDUINO NANO + Esp32 would be very interesting. Without ESP there are many videos, but a video with esp32 well explained does not exist (electronic scheme + ino code, etc). I think many would be interested to see such a video.

[@AndreasSpiess]

Thank you. Unfortunately, I am not too much into LEDs. Maybe other channels will do such a video.

[@tonysfun]

Great video again Andreas! My friends and I love your videos and how you test everything, before you state something that isn't right! Merry Christmas and HAPPY NEW YEAR to you my friend!

[@AndreasSpiess]

Thank you! Same to you! (BTW: I am never sure if I get it right)

[@tonysfun]

Most of us are never sure, we aren't perfect and someone is sometimes smarter than we are in that one or two occasions! But when someone corrects us, we and the rest of the subscribers learn something new - that is always great! I must learn something new every day, that is my life's motto!

[@PhG1961]

Awesome video ! I didn't know much about Neopixels until now ! Thnxs for sharing, and of course congratulations with your channel and the more than 100K subscribers. I hope we all last till your display needs another digit....

[@AndreasSpiess]

Thank you! 900'000 need much more time than 100'000. More than 5 years at least. And then I am 67 ;-)

[@PhG1961]

@@AndreasSpiess By then I'll be 62, but let's face it, makers/creators never grow old.... I hope to go on in my workshop untill I'm 100. I need this, it's in my genes and it would't supprise me it's the same with you ;-) Btw, happy holidays and best wishes for a happy New Year !

[@AndreasSpiess]

We will see. My father was quite healthy up to 90. Merry Christmas to you, too!

[@danielsmullen3223]

Great video. A really nice follow up could explore how many pixels can be addressed before the signal degrades (or you run out of address bits?) and maybe you can give some tips about how to power a giant theoretical display. It would also be really cool to know how fast we can pump data through this thing... can we play a video on a really large panel using an ESP8266?

[@danheidel]

There's no theoretical limit to the number of pixels. The data signal on all these chips is consumed and then regenerated for downstream transmission at each chip. The entire set of pixels is basically a shift register that you're pushing data down. When you send the appropriate signal, all the pixels change their emission to match the shift register data they individually contain. The only real limitation is two-fold. First, each microprocessor can only push out the data signal so fast. At a certain point, your refresh rate starts having to drop to have enough time to run the data down the entire array. With an 8-bit micro, that tends to be in the several hundred LED range and with something like the Teensy's ARM proc, it's a few thousand. The ESP32 should be somewhere in the middle of that, though I haven't tested it. The second limitation is that in my experience, the 2812 chipset is not very reliable and tends to have hardware failures. Shorter lengths of pixels being fed in parallel reduce the probability of a pixel failure in any given strip, so you lose less of the total display area in case of a bad pixel as well as simpler power wiring issues.

[@danheidel]

As for power, these can be very power hungry as you go up in size. The last installation I worked on had something like 30 strips, each 30+ feet in length. It required a few hundred watts of input power when all the LEDs lit up. First, I would highly recommend you use the 12V versions of these strips. That will more than halve the amperage required. The data lines will still be only 5V, so sontrol remains unchanged. Second, I would recommend feeding power in through large gauge wires and powering the strips from both ends to deal with voltage drop across the strips. In that big installation, I made custom boards which had a bunch of electrolytics right by the strip connections to help smooth out voltage drops with sudden intensity increases. I installed power busses that ran up both sides of the array so that there were 14 gauge wires feeding both ends. Third, put your controller on a separate power supply. We had two big dedicated 12V supplies feeding the array and a little SMPS wall wart powering the computer and micros running everything. (don't forget to make sure all the PSs have the output grounds connected to each other - forgetting to do this will make the array not work and might fry all the control electronics if omitted)

[@danheidel]

And lastly, no an ESP8266 won't be able to do a big video display. The installation I described had an Intel Atom based x86 computer to handle the video generation (we were using a rather computation-heavy procedural generation system). Downstream of that, we had a pair of Teensy boards that did the actual timing and bit-banging. The Teensy 3.0 boards have a pretty powerful ARM micro on them that can handle a few thousand pixels reliably. Be user to use the OctoWS2811 library, as it have much higher performance due to some clever DMA stuff it does. I'll be perfectly frank, if you want a really big display, you'll be better off just using a large screen LED TV or projector. The cost per area for the discrete LED strips or panels is pretty high when you take into account all the power supplies and control boards, etc. If you are comparing to a projector, the projector wins hands down on cost. You can just edit the video you want to play to pixelate it and put a diffuser screen in front of the TV/ projection surface. And then, when you get tired of the display, you still have a perfectly good TV. These strips are fun and definitely have use cases such as curved surfaces or outdoor installations or confined areas a TV or projector doesn't work in. But I'll be frank, for a lot of situations, just going with consumer grade gear is cheaper and a lot simpler.

[@dosgos]

@@danheidel Super comments - thank you so much for taking the time to post.

[@zephsmith3499]

Dan's advice is pretty good, but there are some details that could use more elaboration. There is no limit to the number of pixels that can be daisy chained except that the refresh rate drops. These pixels have a constant data rate (effectively 800 Kbps plus or minus some margin that's chip dependent). That timing is the same whether the WS2812b chips are being driven by an ATtiny or a supercomputer. And since each chip needs 24 bits of data (8 bits for each color), that means each chip requires serial output of 30 uS of data (24 bits @ 800Kbps). So 1000 chips requires 30 mS to send data to, and 10,000 chips in series requre 300 mS. (There is also a fixed time needed between refreshes, so the chips will recognize a new frame of data; but the driving code may need that or even more time to generate the next display). The limit on pixels driven by an Arduino UNO is mostly about needing enough RAM to hold the data (along with other RAM needs of the code and libraries). That is, it sends the data at the needed (fixed) rate, but can only do that for pixel data already in RAM. A Teensy (for example) has much more memory and can support longer neopixel chains. More importantly, it can continuously refresh the neoPixels from memory while having CPU cycles available to do other things, while the UNO cannot do anything else while it's shifting out the serial data or it will mess up the 800 Kbps timing. This is because the Teensy can use some DMA tricks, as can the ESP, to send blocks of data to the neopixels without CPU involvement for each chip. (The Teensy can also support up to 8 neopixel chains in parallel). So it's hard to get a large display of neopixels to display video. The fixed serial speed of their data line limits how many chips can be daisy chained before the delays are too long. That can be handled by feeding many separate but syncrhonized pixel chains in parallel. As mentioned above, the teensy can drive up to 8 pixel chains at once, and some designs (eg: using FPGA's) can feed even more pixel chains. You also will find that you need to remap the intensities, as the linear rendering of brightness won't work well for video; at minimum some kind of gamma curve is needed. And finally, be aware that the WS2812b achieves the variable brightness (of each color) using PWM, but the PWM frequency is only about 400 KHz. That is, the chip would render a 20% brightness (value 51 decimal) by being on for about 500 microseconds and off for about 2000 microseconds, for a total cycle of around 2500 uS. The cycle time is fixed at around 2500 uS, and the data determines how much of that time is "on". The cycle time is approximate as each chip can vary somewhat, from other chips and over time or temperature changes, and there is no software control over that. This is relatively slow PWM (some of the other pixel chips have faster PWM; the APA102c Andreas showed is one of the fastest). Anyway, video could potentially show some glitches if your video refresh rate is in the vicinity of the PWM rate without being synchronized. The APA102c chip is preferred by some for video because it not only has a very fast PWM rate (minor help), but it can be updated at far faster then 800Kbps. The 8806 based pixels referenced by Dan also have higher PWM rates and much faster serial update rates, compared to the neopixels (WS2812b chips). They use a 4 wire system (vs the neopixel 3 wire), with power, ground, clock and data. Video on pixel LED chips is a somewhat advanced subject.

[@EvileDik]

The fastLED library has a current limiting function which is nice to use with USB powered projects. I make a lot of nodeMCU light sculptures taking power from Vin pin, set to 400mA in FastLED with 30 LEDs. One of these has been running off a laptop for a year maybe this is safer than trying to power from VCC. :)

[@AndreasSpiess]

I good idea. I have to have a look at this library again.

[@enjibkk6850]

The precise timing thing scared me from using those LEDs, I decided to go for the APA102 type with SPI interface

[@danheidel]

Having done some professional work on large scale installs with the strip versions of these, I would recommend avoiding the 2812 chipset. The timing issues aren't too big of a deal as long as you can tweak the software. However, the 2812 is incredibly unreliable over time. The older 8806 chipset uses a data and clock line in a clock-driven serial approach that is super reliable and seems to be far less prone to failure than the 2812. I haven't used the APA102 but a quick look at the datasheet seems to indicate they are pretty similar to the 8806.

[@foo0815]

Had the same experience, APA102 is much more reliable and has higher PWM rate than WS2812. But the data format is slightly different to 8806, it has an additional global brightness field for all 3 colors, which can be used to improve color resolution or blinking etc.

[@michalsimanek6988]

This is amazing! How did i not know about this stuff yet? I need to get 4 of those panels and make a Snake ripoff or something. Subbed

[@AndreasSpiess]

Welcome aboard the channel!

[@TheRealBobHickman]

There was a batch of those neopixel matrix boards with incorrect silkscreen, so in my case the DIN and DOUT were swapped. Took me quite some time to realize the problem. Always worth double-checking before making something permanent.

[@AndreasSpiess]

Or use Wire Wrapping;-). Thanks for the info. Maybe it will help somebody with the same batch.

[@TheRealBobHickman]

@@AndreasSpiess Have you done a wire wrapping video yet? - I would be very interested to see that.

[@AndreasSpiess]

Yes, a few weeks back.

[@TheRealBobHickman]

@@AndreasSpiess Found it - kzbin.info/www/bejne/gl6XZ2asmM2KZtE - Thanks! Not sure how I missed that one.

[@csongorvarga]

I used Neopixels with Wemos D1 Mini. You can pick different fonts in the sketch. My sketch uses both 6x8 and also 5x7 digits. Also did you notice that when the turn down the brightness the colors start to bleed?

[@AndreasSpiess]

I did not see too much "bleeding". But I did not do comparison tests either. I dim all together.

[@RoterFruchtZwerg]

I've never seen WS2812Bs that initialize full power... good to know they exist! Regarding 3.3V / ESP8266 I've had some bad experiences. I've done lots of projects with ESP8266 and WS2812B and on my bench all worked fine. After having made my own PCBs, they failed... Problem is, the 3.3V logic level is right on the edge what is acceptable for most WS2812B. If your logic level is more like 3.1V OR your Vcc is more like 5.2V the WS2812B LEDs will fail / flicker / show random patterns... You should test your project with slightly higer Vcc which could also happen if you change your wall adapter... The way I solved the problem is by using a Diode between the GPIO and DIN and also pulling DIN up to Vcc. This shifts the logic level up by ~0.6V. With GPIO at GND level, DIN will be at ~0.6V and with GPIO at 3.3V, DIN will be at 3.9V. Also keep in mind that only the first WS2812B is affected, because it will amplify the logic level for all others. You could also drop 0.6V of Vcc on the first WS2812B to lower its logic level threshold. If you don't do anything and use WS2812B directly on 3.3V levels with 5V vcc try to, at least, choose the first LED based on some tests.

[@AndreasSpiess]

Good idea with the diode. I thought, I just add a level shifter if I have problems. I have a few somewhere in the lab.

[@hanpahomeautomation4101]

No problem connecting VCC from ESP8266 to the VCC pin of the NeoPixel if you only have a few pixels like a NeoPixel Stick, individual pixels or only a cut out part of a long strip. So the "never" only applies to when using many pixels. Also, have powered the pixels both from 5V and 3.3V, from the Wemos D1 mini internal power supply, using only a few (16 at most) pixels. I'm using a resistor of 120-130 Ohms between the ESP GPIO and the NeoPixel input and also a 1000uF capacitor on the power supply to the first NeoPixel. Without the resistor, it was not possible to get the NeoPixel Stick working without a level shifter. Using a FET is a good idea but maybe one for each matrix or perhaps 2 and 2. Regarding the decreased intensity to the right, why not just wire VCC/GND directly to each matrix instead of using the daisy chaining? You have room for individual wires.

[@AndreasSpiess]

You are right. And I do not know how much power the LDO chip on the Wemos can deliver. If your 16 pixels are on for a longer time it might get critical. The cap cannot help then, it only helps during short peaks.

[@jmaguilarr]

As always, a fun video , thx maestro

[@AndreasSpiess]

Glad you enjoyed it

[@g3i0r]

Interesting idea with the power FET. For extended panels, it might be a good idea to "gate" every panel individually, so they can be turned on one after the other, like so: power on panel 0 - switch off panel 0 --- power on panel 1 - switch off panel 1 --- ...

[@AndreasSpiess]

Or maybe always 5 together. The voltage drop is only about 5 ms. But it increases with the number of LEDs because it takes longer to transmit the signal.

[@joinedupjon]

@@AndreasSpiess or invent superneopixels with an 'LED enable/disable' input so you can shift in the off states before the full current draw of those eye roasting LEDs clobbers your uC

[@rukminikrishna8162]

God. Will definitely bless you to achive what you desire. My prayers are with you.. Guy with swiss accent. Thank your very much for sharing your knowledge 🙏.

[@AndreasSpiess]

Thank you! 😃

[@DonovanMagryta]

Hi Andreas, What wiring if I want to power the esp32 with the same power supply? My LED strip is 144 pixels so I calculated I will need a 7.2 amp 5v power supply but can I connect that to V-In?

[@AndreasSpiess]

You can try. But microcontrollers are picky if it comes to voltage drops...

[@DonovanMagryta]

@@AndreasSpiess Thanks for the tip. I ended up going with a 5v 2a power supply and using only 72 pure red green or blue WS2812B LEDs. That way it only pulls 20ma per pixel. To avoid overcurrent on the MCU PCB traces and the builtit zener diode, I am powering both the ESP8266 MCU and the LED strip in parallel. MCU 5V(USB)----5VCC 2.1A-----LEDV+ MCU GPIO4----LED Data In MCU Ground----5V 2.1A Ground-----LED Ground.

[@MkmeOrg]

Great stuff Bud!

[@AndreasSpiess]

Thank you. And Happy New Year to you!

[@TMS5100]

i prefer apa102/clones. they are more tolerant of timing and voltage levels. also capable of much higher refresh rates than ws2812. also led defaults OFF. @10:06 wemos d1 mini is optimal size for esp8266 since most pins are wasted anyway. d1 only exposes the useful ones.

[@AndreasSpiess]

I had to take whatever was on the panels :-(

[@rogeriorogetheo6202]

Good afternoon, Andreas, I'm trying to use Jinx / Gladiator with PES 32, but I'm in trouble, I could control myself.

[@kenheitmueller69]

That's a lot of current for that thin wire wrap wire. you'll have less brightness uniformity issues if you use heavier wires for GND and 5V for your panels. also, you have Neopixel spelled "Neopixle" in your slide at 1:19. Thanks for your videos as always!

[@AndreasSpiess]

Because I only have about 500 mA divided by 6 panels I have no issues with power. But if you want to have all LEDs on you are absolutely right.

[@katelynch701]

Hello I was wondering if you could help me : I want to program a strip of neopixels to turn on and off in at a random point in a given time intervals. Do you recommend any libraries to achieve this ?

[@AndreasSpiess]

I would use Google to search for such projects. You will find many.

[@Pluto1010]

Why are the WS2818 enabled on power on? Wouldn't it be better that they start powered off? I can tell which one has to get bright.

[@AndreasSpiess]

There is no need and I think, this is wrong. But I do not think it is written in the specs.

[@science4allworld587]

Looking forward for 999999999999 digits soon :). Great job as usual.

[@AndreasSpiess]

:-)

[@satviksharma1146]

Can't we use decoupling capacitor to prevent esp8266 from crashing?

[@AndreasSpiess]

You can try. With some love and effort it will work ;-)

[@kanapkazpasztetem]

I like that you started using this plastic pointer - it makes look videos more professional imho :D I saw that you were using LDR for light measuring, could you maybe someday make a video on light measuring? (LDR vs photodiode vs other sensors like TSL2591 that will return value in Lux)

[@AndreasSpiess]

I have a few sensors and the topic is on the list...

[@superdau]

If you want even brightness, you should connect + to one end and - to the other end of the strip. Connecting in the middle just halves the orignal problem ;) .

[@AndreasSpiess]

Interesting idea. But accurate. You are right!

[@NiHaoMike64]

Add a resistor across the FET to give the LEDs enough power for the logic to work. Then you can command them off before enabling full power.

[@AndreasSpiess]

I do not think this is a good idea because the resistor has to be nearly zero ohms to get enough power to the LEDs.

[@xz86]

hello great video. i have a question when it come to printing color to each LEDs. i have 50 LEDs in a strip. right now, my code has 50lines for each leds. is there a better way to print led? 0-9 to be red and 10-19 yellow and 20-50 green. i try for loop, but it didn't seem to work too well.

[@AndreasSpiess]

There are libraries for that.

[@AdityaMehendale]

Ouch! Andreas, I think the open (untwisted) wiring at 9:44 will cause problems in the long run. If you just keep a 3x-braid, and daisy-chain (no ground-loops) it'll be more robust and less noisy.

[@AndreasSpiess]

It is not a life-threatening device. So I can change it if it starts to show the wrong numbers. So far it only could advance a little faster. But this is not because of the wiring, I assume ;-)

[@AdityaMehendale]

@@AndreasSpiess 105k is fantastic :) Besides - the quality matters more than the quantity, right?

[@AndreasSpiess]

I have a very good quality of subscribers. I always see it if I have a successful video. Then, the quality of comments decreases because non-subscribers also comment.

[@HeathenHacks]

I used a 5v 40a power supply for my 6 panels of 8x8 neopixels on an arduino uno with a 470ohm resistor connected on the data pin. Also used the same library examples and some of the examples gives my panels ghost pixels. I'm not sure if my power supply is the problem or if I am missing something. Can you guys help me out? Thanks!

[@AndreasSpiess]

Maybe you also have a batch of panels which have DIN and DOUT mixed up? Some other viewers got such panels.

[@HeathenHacks]

Andreas Spiess thanks. but, no, i don’t think so. Now my problem is I think I broke the DIN on my other matrix. lol. The 1st column on that particular matrix is the only one working now. 😂😂😂

[@spikeydapikey1483]

Good tip about powering from the middle. Not heard that one before :o)

[@AndreasSpiess]

:-)

[@thedog2320]

hey there, question for you. i need to power 15 neopixels but just the green, or the red. Is it ok to do so just using a arduino nano or micro? if i understand correctly that should put me in the 300 milli amp range which is bellow the 500 cap right? is that ok or should i get an external power source?

[@AndreasSpiess]

Maybe you look at the datasheet of the Neopixels? I do not have such numbers in my head. The nano and micro have a direct connection from USB to the 5V pin. So you just need a strong USB charger.

[@babakhallajian684]

Hi Andreas Is that possible to use neopixels as a ceiling light with 1800 lux? and how much power consumption in comparison with normal LED light? Is it worthy? Thanks

[@AndreasSpiess]

I never did any lux measurements, but I assume you get bright light if you add many strips in parallel. You find many videos creating all sorts of configurations with Neopixels. The power consumption should be similar as Neopixel use the same technology as normal LEDs

[@babakhallajian684]

@@AndreasSpiess Thanks Mate

[@michaegi4717]

Good video! Thanks for all the information.

[@AndreasSpiess]

You are welcome!

[@maxbarbul]

Thank you for your video, Mr. Spiess. I was struggling to power ws2812b strip with 18650 batteries. I had strange effect, i had to touch the batteries to make strip light. Connecting ground of strip to d1 ground solved it. When powering with usb buck-boost converter there was no issues. Could you explain why it’s important?

[@AndreasSpiess]

I do not know fron the distance :-(

[@maxbarbul]

Andreas Spiess no problem, you helped me anyways :)

[@sdrshnptl]

amazing elaboration!👍

[@AndreasSpiess]

Thank you!

[@Davedarko]

Congratulations for hitting 100k subs :)

[@AndreasSpiess]

Thank you! Home again?

[@Davedarko]

@@AndreasSpiess Yup! and already preparing for chaos communication congress :)

[@driftlessheights6177]

Hi Andreas, Just found your channel and loving it. Would appreciate your advice. I am finishing my basement / movie room finally. Planning on integrating animated LED stripping into the crown molding around the room. 80 total feet / 25 meters. These Neopixels look good and my Arduino / Raspberry skills are sufficient. However I looked at the the pricing of the Neopixel strips is higher than the other color strips. At 25 meters the prices on Adafruit are $20-$30 per meter cost would be $500 to $700. Is this overspending? Is there a less expensive but acceptable alternative? Second question is related to wiring. I know we can minimize power voltage drop by parallel-izing the 5v and ground. But is there a voltage drop around the signal that we need to consider at 25 meters? What kind of wire would be needed over that length? Love your work!!! Tom

[@AndreasSpiess]

1. Adafruit is usually more expensive than AliExpress 2. Data line has no loss because it just goes from one pixel to the next. But you have to check the maximal number of supported pixels. Maybe you google for similar projects.

[@driftlessheights6177]

@@AndreasSpiess Thank you Andreas. I will take your advice and try the other purchase source. Perhaps the individual LEDs dont need to be so close together for my project and that should drop the cost also. By the way, have you ever done a video on chosing wires for projects based on maximim estimated aperage? I have been stripping wires from ethernet cables for some of my projects but dont know how hard I can push them. I know stranded and solid wires are also factors, besides wire diameters.

[@florianzschocke4253]

Keep in mind that if you have a lot of pixels, the simple Arduino, like Uno or Nano, with their 2k RAM will not have enough memory to program the pixels. You will need a MCU with a bit more RAM if you have more than a few 100 pixels.

[@jeremyviny6817]

Hi, I have been trying to use 8 neopixel matrices (8x). For some reason anytime I have more than 4 panels it stops working? For Context** I am controlling the neopixels with a joystick. Anytime I upload a sketch with more than 4 panels. The LEDs do not reset to the start position and I am unable to control them with the joystick. I am big fan of your videos! Thanks

[@AndreasSpiess]

Unfortunately, I cannot do remote debugging. Maybe you find a project with a similar setup of the neopixels and you can check them out with a different sketch.

[@ghanrajsingh7347]

As ALWAYS AWESOME VIDEO AND INFORMATIVE I LEARN A LOT FROM U R CHANNEL

[@AndreasSpiess]

Thank you!

[@snipe2701]

over time did you run into trouble with the FET going pop? or is it still working?

[@AndreasSpiess]

I do no more operate it. KZbin does no more show the actual numbers :-(

[@dannyarnold9823]

Could you use a large capacitor to delay the start of the matrices instead of the pfet?

[@AndreasSpiess]

I do not think this would work.

[@paulvandenbossche]

Then you need an even bigger Power supply to charge te capacitor ...

[@jonathanmegevand3266]

Thanks 👍 great video once again !

[@AndreasSpiess]

You are welcome!

[@aakash512]

Hi Was just wondering if we could use a capacitor with suffeciently large discharge time for startup of the display?

[@AndreasSpiess]

A discharged capacitor dis like a short circuit in the first moment. It only would help with a FET. Then you could charge it with the display off

[@aatkarelse8218]

What a nice project !

[@AndreasSpiess]

Thank you!

[@dominikworkshop6007]

great video! :D But I would connect +5V to one side of the display and ground to the other side to to make sure that all panels get the same voltage instead of connecting power to the panel in the middle

[@AndreasSpiess]

You are right. But you need thick cables. For me the middle was very handy because I had a connector right there with thin wires.

[@xyloidify]

Another wonderful video, thank you!

[@AndreasSpiess]

You are welcome!

[@masami1986]

What happens if you set the color of a pixel and then disconnect data pin, its still on with the color setted?

[@AndreasSpiess]

No