Meanwhile the printed circuit boards arrived which I have ordered from Aisler based on my first draft. The quality is fine. There where only production-related burrs that I could easily sand off.
A first test was also successful:
The power requirement is – depending on the brightness setting of the displays and the displayed content – about 2 and 5 watts. A test with maximum load, which means all LEDs activated and with maximum brightness, still has to be done.
During the test I noted that the Raspberry Pi Zero starts making a clearly audible whining noise when the 5V output gets loaded. This was surprising as I assumed that this wire is connected directly to the USB port. Maybe I need to find another solution for this.
Another unexpected effect: when powering on, the GPIO ports of the Raspberry Pi produce a signal which causes the displays to show undefined content. First I thought there was an error with the reset circuit – but even when manually connecting RST with ground, the display only stay blank as long as the reset input is active. As soon as the reset input gets deactivated, the random output appears again.
Even if not everything is perfect yet – for a first try the result is already very promising :-).
Fixing the whining noise
The whining noise I experienced when many LEDs got activated at the same time was propably also caused by the clock generator in the display modules running at 57 kHz and the LED matrix which gets updated 256 times per second.
To fix this I connected a filter capactior with 100 μF to each display directly at VDD and GND. Most likley a much less would be fine as well but I had these capacitors as spares and mechanically they also no problem. The effect could not be eliminated completely but it is much weaker and barely audible. I think when I mounted the construction into a closed case I won’t hear anything at all.