WS2812/WS2813 single LED chips: Resistors? Data Channel voltage drop?

Other Hardware LEDs and Multiplexing
1

Hey everyone,

i am more a chemical guy, but i want to build a quite large LED Cube. So I had a few questions, i would be very grateful if someone might answer.

1496 RGB LEDs
50fps
12m³ cuboid
non regular arrangement

I was searching the web and i think a teensy 3.6 + octows2811 is a cheap and reliable base.

Two questions:

  1. Since i need to solder every LED anyway i thought i could buy the chaper LED chips alone:
    https://de.aliexpress.com/store/product/10-1000pcs-WS2813-6pins-Dual-signal-5050-SMD-Black-White-version-WS2813-Individually-Addressable-Digital-RGB/1051119_32770287791.html?spm=a2g0x.10010108.1000023.22.20eb376cPbucVT

The problem is, some of the presoldered products have an extra resistor (i think) on the pcb. Is this necessary or is something like this ennough:

1354×458 77.7 KB

I can't access broken LEDs very easily, so how do I reduce chances to harm the LEDs in general?

  1. For the 5V wiring I know the voltage and ampere and therefore can calculate the mm² of the wire.
    But i have no idea how much ampere and voltage the data pin uses, so i don't know how I can calculate the required diameter.
2

I would not recommend using SMD LEDs for building a cube.
The SMD parts are usually not very stable mechanically, therefore your cube would probably end up being pretty fragile. Also, since there isn't any real housing on those LEDs, the light emitted will be difficult to see from the side.

You might want to consider using WS2812 LEDs in a frosted 5mm housing.
Even with those, it is still a difficult and tedious task of soldering the leds because you need to split the wire carrying the data signal at each and every LED.

The Data pins use 5V, but there's only very little current flowing though them, probably somewhere in the µA range.

That "protection circuit" should not be necessary when you power everything from the same power supply.

You will need to test everything after every step of the assembly so you can spot faulty LEDs before your cube is completely assembled. Usually when building a cube you would first solder rows/colums, then test everything, then assemble those into planes, test everything again and finally assemble the whole cube and test it.

Question from me: What diameter of wiring are you planning to use?
1500 WS2812 LEDs will draw around 100A of current when all channels are at 100%.
You are going to need some VERY serious wiring for that kind of current and the power supply will also not be cheap.

3

Build a small prototype first. Learn from your mistakes before you waste a lot of time and money. Maybe start with 4x4x4 cube?

The small component on the pre-soldered boards is probably a cap, not a resistor. Maybe 0.1uF.

4

It will only be possible to view the cube from beneath, so the 120° emission is not a problem.The frosted housing is great, but the wires are really thick. Should be as thin as possible, as thick as needed, I also wouldn't use any isolation, but transparent glue every few cm to keep the wires apart, so i probably need a fuse?
I will use transparent nylon wires to space the LEDs correctly.

I thought i can drown the middle of the wire in solder of the first LED and connect the end at the alternative input pin of the following LED, so no splitting needed.

I wanted to use 2 x 5V 60A power supplys.
https://www.ebay.de/itm/DIY-Schaltnetzteil-5V-12V-24V-48V-Netzteil-Trafo-LED-Strip-Power-Supply-Free-DHL/152748036902?hash=item23907e1326:m:myTOdGzfLZ4EoPH-d9iEjWg

Placing the supplys at the top and only connecting strings of max 15 LEDs to the bottom. So running around 100 stripes in parallel at the 5V supply and 187 LEDs as one data stripe, using 8 data stripes total.
4,5W maximum power draw, 3m copper wire, 1mm² should result in less than 4,9V at the end of the cord. I read somewhere, this is ennough to run the LED.
If the power at the data pin is really that low, then the 30m of 1mm² wire for one data stripe is also ennough to stay above 4,9V.

I also didn't solder much before so this is another problem. Copper can't easily being soldered, than i would need to use iron i guess? I don't know how much voltage drop is okay, so in this case I would need to use a thicker cable.

I planned to order a few more LEDs, so i can try the soldering, the wiring and the controller, before i start the bigger project, so i totally agree.

The cap is probably used to maintain the 5V if something results in a voltage drop, so the LED is running even if the wires are a bit too long or the supply can't deliver constant 5V? So is it necessary?

5

1mm² wire can run several amps without any problems. For the data connections I would use something like 0.2mm² or even thinner...

However, I would really really suggest you go with @PaulRB 's suggestions and try out a small 4x4x4 Cube first.
You most likely will notice problems that you never thought about (although I admit you seem to have put quite a lot of thought into your project already).

6

  1. Data in of each LED must go to Data Out of the next one. It sounds like you think you can just run one line along data in - that would send the same data to each led, not what you want.

  2. Viewing angle of the SMD leds is kinda narrow, and they don't look good at the edges of that range of viewing angles either. They also look kinda shitty without some sort of diffuser - the points of light are really small, you can see the individual leds inside the WS2812, and their relative brightness shifts a little with the angle. If your viewing angles are restricted enough that you don't care about the viewing angle, why bother making a cube?

  3. At 800khz, you can transfer 100,000 bytes of data per second. 33,000 LEDs. You want to use ~1500 LEDs, but you want 50FPS. With 1500 leds, you could only get a maximum of 22FPS. You need to use fancier APA102 LEDs or similar if you want to run faster (and they have both a data and clock line, so more wiring to do)

  4. The through-hole WS2812B-alike clone APA106-F8 is a frosted through-hole WS2812.

  5. WS2812 needs a 0.1uf ceramic cap between Vcc and Gnd right near the chip for decoupling, otherwise it will turn itself off or reset randomly during operation.

Wiring that up with SMD chips sounds like hell...

You should definitely make a smaller prototype.

7

Copper can't easily being soldered,

No copper easiest to solder, iron is very difficult.

Search for @LMcUnderwood on twitter for her very nice way of making an LED cube.

8

Shuzz:
However, I would really really suggest you go with @PaulRB 's suggestions and try out a small 4x4x4 Cube first.

I will do that.

DrAzzy:

  1. Data in of each LED must go to Data Out of the next one. It sounds like you think you can just run one line along data in - that would send the same data to each led, not what you want.

I was planning to use the ws2813. Maybe I got it wrong, but there I thought the wire that is going to the first LED's data-in has to go into the backup data-in of the following LED, so it can keep on running if the first is broken.

DrAzzy:
2. Viewing angle of the SMD leds is kinda narrow, and they don't look good at the edges of that range of viewing angles either. They also look kinda shitty without some sort of diffuser - the points of light are really small, you can see the individual leds inside the WS2812, and their relative brightness shifts a little with the angle. If your viewing angles are restricted enough that you don't care about the viewing angle, why bother making a cube?

It is more like a sky, going 2 meters high with a mirrored ceiling. Lying under it i hoped it wouldn't be a problem. But this I will see with the small prototype.

DrAzzy:
3. At 800khz, you can transfer 100,000 bytes of data per second. 33,000 LEDs. You want to use ~1500 LEDs, but you want 50FPS. With 1500 leds, you could only get a maximum of 22FPS. You need to use fancier APA102 LEDs or similar if you want to run faster (and they have both a data and clock line, so more wiring to do)

I thought this is per channel. If I use a teensy 3.1 or higher, I have up to 8 channels per board, so I only drive 187 LEDs each. Thought this would solve the fps issue. Otherwise i buy two boards and stay with 44 fps.

DrAzzy:
4. The through-hole WS2812B-alike clone APA106-F8 is a frosted through-hole WS2812.

Is the APA106-F8 using the same protocoll like the ws2812, so i can stick with the teensy board? Because especially with the cap in mind, it isn't such a great idea to solder all the chips.
On the other hand, why aren't there any caps on some of the pcbs of 2813? Or do I just don't see them?

Grumpy_Mike:
No copper easiest to solder, iron is very difficult.

Search for @LMcUnderwood on twitter for her very nice way of making an LED cube.

Thank you.

9

Out of curiosity: What dimensions will your finished cube have in terms of LEDs?

8x8x8? 10x10x15? Something entirely different?

Because, honestly, maybe (just maybe) it would really be easier to build and run by using the "traditional" multiplexing, maybe coupled with some constant current drivers like the TLC5940.
Don't let the lower brightness scare you off - it is actually not that bad in reality as long as you do a 1:8 or say a 1:10 multiplex the LEDs will still be bright enough. Also, if I understand you correctly, you're not trying to illuminate your room but rather make a fixture that you can look at while lying on your bed, right?

10

This is what it should look like, when it's done.

I looked for some multiplexer tutorials. The ws2811 LEDs provide a much easier handling for anyone who only has school tought knowledge about electrics.
What is the advantage of a multiplexer setup?

Prototype.gif

11

Prototype.gif

12

I think that multiplexing 1500 RGB leds would be a nightmare. Multiplexing could be done on the prototype at 4x4x4 or even 8x8x8 but then it would no longer be a true prototype for the larger cuboid.

13

What is the advantage of a multiplexer setup?

Only cost, and you swap that for a whole heap of other problems.

14

About the maximum fps with a teensy board, does someone know if it's for one board or one connection per board?

15

Per board. The frame rate is determined by how fast the data can be sent to the LEDs so it is the same for all boards irrespective of the speed of the processor. This is because the faste processors need to be slowed down to keep the data going at the rate the LEDs demand.

16

Okay, so I guess I need the octo Adapter:
https://www.pjrc.com/store/octo28_adaptor.html
There they say over 4000 can be used at 60hz.