Hello. I have an idea to assemble something like a screen of 100 controlled LEDs. SMD 5v RGB WS2812 size 2020. https://aliexpress.ru/item/32876876257.html On the Internet, in most cases, each LED has a capacitor. I have no place to install them. Is it possible without them? And what will be worse in this case? What will be a plus when installing them?
Is it possible without them? And what will be worse in this case? What will be a plus when installing?
All digital circuits generate electrical noise in the power supply, including the various types of controllable LEDs. The capacitors stop (or decouple) the noise from one component interfering with any of the others. Most likely if you don't include them you will get all sorts of strange effects and problems with no obvious cause.
With an array that big it is best to fit a large capacitor as big as you can get at the start and maybe even the end of the chain.
Like Peter said it is to help to remove sudden changes which cause interference. This topic is known as decoupling and I write a small page on the what and why of it.
http://www.thebox.myzen.co.uk/Tutorial/De-coupling.html
AdaFruit sell a lot of different devices this is a guide to how to treat them
They are cheap for a reason you know, not the best quality products you will ever find.
You definitely need the small decoupling caps along the chain - sorry. A big cap at the ends won't cut it.
Without the cap, they can reset when you change the brightness from dim/off to something bright. Same reason all digital IC's get a decoupling cap. The WS2812 has a digital IC in it, but it doesn't have an internal cap; you need to provide that, and it needs to be right next to the chip. You can get away with one cap for a few nearby LEDs (I - and adafruit - did that for a circular WS2812 array*), but you can't omit them completely.
- I took adafruit's design and added one more ring of LEDs around the outside, iirc
You definitely need the small decoupling caps along the chain - sorry. A big cap at the ends won't cut it.
Yes this is taken as read. Those Adafruit matrixes have these but also recommend a large cap as well. If you don’t have both then you can burn out the small ceramic ones.
The thing I was trying to point out is that given the source of this matrix they might have no ceramic capacitors. Some strips I have seen only have ceramic capacitors every 16 LEDs, so not good.
Grumpy_Mike:
If you don’t have both then you can burn out the small ceramic ones.
What do you mean by that?
I mean what I say. With no large bulk decoupling capacitors the ceramic capacitors can burn out because the ripple is such that they are trying to pass too much current and they end up burning out.
Back in 2016 I was making a 3D printer, using good quality power supplies and Pololu stepping motor drivers.
The ceramic capacitors on the drivers burned out because I had no bulk decoupling apart from those in the power supply. Fitting bulk decoupling cured the problem, and the driver boards were repaired by replacing the ceramics.
Given that driving 4 stepping motors is about the same sort of load as a decent length of WS2812 LED strip the same thing could happen without bulk decoupling capacitors.
Grumpy_Mike:
Some strips I have seen only have ceramic capacitors every 16 LEDs, so not good.
Well, if THOSE work, that gives one an idea of what you might be able to get away with...
Also, ooooh, those are the super-tiny ones!
And something tells me you're planning to put them as close together as you can
Maybe tiny caps on the backside, with a via for each LED connecting to them?
Grumpy_Mike:
The ceramic capacitors on the drivers burned out because I had no bulk decoupling apart from those in the power supply. Fitting bulk decoupling cured the problem, and the driver boards were repaired by replacing the ceramics.
I am reasonably sure the capacitors you cite were not the 100 nF ones were are talking about here. They would have been at least 10 µF, one hundred times the capacity and therefore passing 100 times the ripple current. 
They would have been at least 10 µF,
No but probably about 2uF.