Data point. These 8 x 16 pixel panels:
… pull about 2.25A @ 5V on full-white according to my multimeter. Which is about 18mA per RGB LED which is a lot less than I thought they would. Aren’t they meant to be 20mA per R, G and B, i.e., 60mA per WS2812B?
Or am I just about to start smelling something very funny from the (2.1A rated) Apple USB wall-wart I’ve got them plugged into?
I can confirm your numbers with a recently purchased 5m ws2812B strip from Ray Wu on aliexpress: 18mA per LED at full white
Well that’s interesting then. So I only needed a 40W PSU for the 32 x 18 panel I’m building, not the 200W one I bought?
Probably. I reached the same conclusion on a smaller scale and got a 20W PSU instead of 80W. The voltage on my PSU drops to ~4.5V under load. Maybe the 200W PSU will perform better than the 40W when you light up the whole panel?
Also note that you could hook up a power sucking 12,000 BTU monster electric air conditioner to a “2.1 Amp” power supply and you wouldn’t see it drawing much more than 2.1 Amps… because that’s all the power supply can deliver.
I’ve had this problem before.
If you want to measure how much current something wants to draw, you need to connect it to something that can deliver at least that much current. Hook it to a 5v 10Amp supply and see how much current it draws.
The fact that it can operate on less is definitely interesting, too, but I think you might not be headed for happy power engineering land.
As you reach the 90% mark, you will probably see a drop in the voltage from your power supply as it starts to struggle. Plus it may start moving away from clean DC to something with other frequencies in it. If you have a scope, put that on and see what happens.
Most power supplies will allow you over the 100% mark. But my guess is the output voltage will drop and there is a significant chance that longer term the internal heat generated in the power supply will do irreparable damage to the power supply. Good power supplies have fuses of some sort to stop that (resting or single use). Cheaper and smaller volumed don’t (aka wall-warts).
Low voltage causes a multitude of problems depending how far off it is down and the voltage tolerances of everything being powered. The unclean DC will probably also cause data corruption and timing issues.
Heat damage can be as quick as a few seconds to several days, and everything in-between.
I normally use 20mA for the calculations because I can do it in my head. Maths with 18 is much harder 
This excess also gives spare capacity for the power supply to avoid getting close to it’s limits and generating more heat.
Bottom line - don’t stress your power supply. Its good practice and saves disappointment or worse a fire. 80% to 90% capacity is probably a good rule of thumb for a maximum.
Last time I measured (with the tools I had on hand), I got these values for the current draw of the LEDs in a strip of WS2811’s:
20mA is a great estimate for how much power each pixel is likely to draw, presuming that you’re mostly using “pure hues”. If you’re using desaturated (whiter, more pastel) colors, you’re going to go higher.
Also worth noting that the pixel closest to the power supply will use a little more current than the pixel furthest from the power supply due to the voltage drop along the strip. Use the (NUM_LEDS / 2)th pixel for testing, or better yet, use 100 pixels, and divide resulting current draw by 100. (I think that’s what I did to get the numbers above.)
I’ve done some more experiments (for SCIENCE) and here’s what I found:
Hooked up to the beefy 200W supply, the 128-pixel panel draws… 1.96A @ 5.11V (I only have one multimeter so these measurements weren’t taken simultaneously, but I guess this will still be roughly correct?) for a nice neat total of 10W or so.
I hooked it back up to the Little iPad Wall Wart That Could and again got my 2.26-something amps, but when I measured the voltage it went all the way down to 3.2V or so. It didn’t seem much less bright, if at all.
So, we’ve learned five things:
… and finally: