thanks for all the feedback. I've actually tried using external 5V power supply and the voltage drop is the same. I've now finished a 4 layer PCB design with separate 5V & GND layer as well as an option for a separate USB power. Will see if this improves the output RGB signal more.

It strikes me that you have small bypass capacitors in appropriate places, but there is no big bulk power-supply stabilising capacitor next to the power input, where I'd expect to find it. Whenever you have a supply with significant impedance and a load with short-term transients, you'll need that to avoid dropouts.

thanks, I will add a 470uF capacitor next to the transistors.

Oh yes, this.

We learned early on in the Raspberry Pi days that it was a race to the bottom on both "PSUs" and USB cables )-:

-Gordon

_________________
--
Gordon Henderson.
See my Ruby 6502 and 65816 SBC projects here: https://projects.drogon.net/ruby/

To say nothing of outright violations of the spec. A Google engineer engaged in testing cables found out the hard way that one particular USB A-C cable had the power lines reversed (which blew the power input circuitry on his laptop, which used USB-C for charging), and the high-speed data lines were missing entirely. He'd bought the cable on Amazon.

Ditto on that.

While on the subject of bypass capacitors, radial-leaded parts usually introduce less series inductance into the circuit than axial-leaded ones. Series inductance is your enemy when it comes to effective bypassing. I use X7R MLCCs at all active devices, plus an X7R MLCC across the power input jack in parallel with a large, low-ESR electrolytic. The purpose of the MLCC at the power input is to suppress any switching transients that don't get suppressed by devices' MLCCs.

Ideally, each device's Vcc pin should be wired to the device's MLCC and the MLCC bonded to the inner power plane. The device's Vcc pin should not be directly connected to the power plane. The purpose of doing so is to force all switching transients back to the MLCC rather than allow transients to enter the power plane. This is a design technique I learned decades ago and has been advocated by industry gurus such as Dr. Howard Johnson.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

thanks, I will implement this recommendation.