Can you post the entirety of your code? I want to see what you have - also what version of the library are you using, have you tried grabbing the latest version from github?
(It’s possible that the timings in the data sheet are off - but the cycle counting is pretty exact for the output - but I’ll test with a scope later)
Code : https://drive.google.com/open?id=0B6EgmG1E-oOEYUlVYkNHb1FIMkE
(second video)
Other code test (ColorPalette) : https://drive.google.com/open?id=0B6EgmG1E-oOEcDFPMWtHME44S1E
Arduino : 1.6.7
FastLED : 3.1.1
Hi again @CeeeJ , sorry to harass you on this specific issue but please confirm that you have indeed tested your circuit by removing ALL of the 100 Ohms resistors to 5V and not just one to try it quickly.
From the circuit diagram you posted, you have 100 Ohms resistors to 5V and assuming a typical white LED current draw, that would be around 45 milliamps and the voltage drop on that 100 Ohms resistor would be 4.5 V.
The logic device integrated in that LED just cannot cope with only 0,5V to operate.
Even if you never drive the LEDS white, it is obvious to me that any constant current draw from the LEDs combined with the instantaneous current demands of a switching IC could easily explain the erratic behavior that we are seeing in your posted videos.
That does not mean there is not still a timing issue but I strongly recommend, if you have not already done so, to remove all the 100 Ohms resistors in your circuit.
Jean-Pierre
I’m also playing with some PL9823. Those kind of glitches are usually due to GND issues or inconstant / too low power supply. I have a similar problem with my NodeMCU v3 at the moment. Even I’ve connected the GND rail of the LED +5V power supply to the The GND of the MCU with a 10KΩ resistor, I still have glitches as long as there is no WIFI traffic yet to the ESP8266. Resistors are useful only to protect from over voltage on the LEDs. As JP Roy already mentioned it, the 100KΩ is probably too high,. I’m not using resistors on the +5v rail but I use one on the DIN of the first LED chip. A unique decoupling capacitor on the +5v rail (470uf in my case) is enough. You can also add a high speed Zener diode like described on the WS2811 datasheets to prevent noise on the DATA line.
@JP_Roy Well, I just re test and It Works!
You did well to harass me.
I do not understand, I had already tested this circuit.
I am ashamed to have you wasted your time at all.
I am pleasantly surprised all the answers in your community.
Thanks for your help. I can continue my tests.
Can we avoid the blue startup?
Does that blue startup flash happen constantly? Or is it something about the program that you’re using?
Hi @CeeeJ , don’t worry… you did not waste my time . I am happy to have helped and even more happy that you reported back.
Don’t hesitate if you have other issues !
Regards, JP
Hi again @CeeeJ , you can’t do anything with that blue startup… that is a ‘feature’ of those 8mm RGB LEDs that they power-up blue… 
I use the example file ColorPalette.ino (Github). See the video.
1: I plug the power supply.
2: LEDs light up in blue. (Whatever the program).
3: The program starts normally.
So I just want to know if this is normal and if we can avoid it.
@JP_Roy Ok for blue LEDs, it’s not too serious. (This verifies that all LEDs are operational ^^)
Thank you
That’s the normal init for the PL9823. I’ve oredered 128 of those LEDs. 3 came dead on arrival. it means you can expect a good 2% of the ICs to have problems (well consider the PL9823 are very cheap on ebay). A working IC with a problem on one of the 3 RGB connections will probably happen on less than 1/1000 pieces. When you have thousands of those guys to check, it’s way faster than doing an init with the MCU and checking one by one with a RGB test pattern.